Timestamps are in GMT/BST.
[0:21] <jabawok_h2> btw thanks to KenMan, salahx and hirvox for your help, 5090 now running purring along with nptl on gentoo.
[0:22] <salahx> great
[0:23] <jabawok_h2> load is till a little high but thats due to major swappage on a 384M server with lots of other stuff running
[0:24] * salahx (salahx@sc1-24.217.174.147.charter-stl.com) Quit ()
[0:24] <jabawok_h2> i'll throw a bit more memory in , sometime and see what happens
[0:30] * moskau23 (~Miranda@dsl-082-082-224-034.arcor-ip.net) has joined #freenet
[0:31] * moskau23 (~Miranda@dsl-082-082-224-034.arcor-ip.net) Quit (Read error: 104 (Connection reset by peer))
[0:32] * moskau23 (~Miranda@dsl-082-082-224-034.arcor-ip.net) has joined #freenet
[0:37] <Kyon> i get 0 unreachable, 0 restarted, 0 rejected, and 3 backed off. this possibly improve? or is there a problem with my seed nodes
[0:46] * lostlogic (~lostlogic@node-402418b2.mdw.onnet.us.uu.net) has joined #freenet
[0:52] * lostlogic (~lostlogic@node-402418b2.mdw.onnet.us.uu.net) Quit (Remote closed the connection)
[1:28] * oierw (mathew@cpe-69-75-112-240.hawaii.rr.com) has joined #freenet
[1:44] <mazzanet> bring bring
[1:48] <KenMan> ah, hello ?
[1:50] <mazzanet> moshi moshi!
[2:02] <jabawok_h2> mazzanet, out of interest, where are you based?
[2:02] <mazzanet> not japan
[2:04] <KenMan> mazzanet connects through the interstellar links... he is a russian space-dog that crash landed on venus
[2:04] <jabawok_h2> oz?
[2:05] <KenMan> same thing
[2:05] <mazzanet> ya
[2:05] <mazzanet> land of .au
[2:05] <jabawok_h2> wa?
[2:05] <mazzanet> vic
[2:06] <jabawok_h2> ahh. i'm sure i've seen u on #perthwireless or something to do with the wafreenet (a wireless network, nothing to do with www.freenetproject.org)
[2:06] <jabawok_h2> or perhaps a SSID of a wireless network whilst wardriving or something
[2:14] <mazzanet> heh
[3:50] * theLostFloppy (~michaelku@fia41-111.dsl.hccnet.nl) has joined #freenet
[4:46] * Kyon (~Kyon@pcd396122.netvigator.com) Quit (Read error: 104 (Connection reset by peer))
[6:08] * Hory (~Miranda@81.196.25.110) has joined #FreeNet
[6:28] <Ash-Fox> But you see it's not me, it's not my family. In your head, in your head, they are fightin'. With their tanks, and their bombs. And their bombs, and their guns. In your head, in your head they are cryin'. In your head, in your head, Zombie, Zombie, Zombie, eh, eh. Whats in your head, in your head Zombie, Zombie, Zombie, eh, eh, eh. oh, doodoododododoodoo.....
[6:28] <jabawok_h2> classic song.. pity it's their only good one
[6:28] <Ash-Fox> jabawok_h2, listen to depache mode
[6:29] <jabawok_h2> and?
[6:29] <Ash-Fox> depache mode is classic too, and good
[6:29] <jabawok_h2> fair enough
[6:29] <Ash-Fox> Another mother's breakin' heart is takin' over. When the violence causes silence. We must be mistaken. It's the same old team since 1916. In your head, in your head they're still fightin'. With their tanks, and their bombs. And their bombs, and their guns. In your head, in your head they are dyin'.
[6:30] <jabawok_h2> relevance to #freenet? 0%
[6:30] <Ash-Fox> jabawok_h2, I prefer this to kiddiepr0n
[6:32] * jabawok_h2 turns on troll protection
[6:32] <Ash-Fox> jabawok_h2, impossible on freenet, remember! Free speech! :P
[6:32] <jabawok_h2> lol @ impossible
[6:33] <jabawok_h2> ignore Ash-Fox
[6:33] <Ash-Fox> you forgot the /
[6:38] <theLostFloppy> Aaah. Fuqid!
[6:48] * silent-jaymz (~jaymz@dialup-226.82.221.203.acc52-dryb-mel.comindico.com.au) has joined #freenet
[6:51] * theLostFloppy (~michaelku@fia41-111.dsl.hccnet.nl) has left #freenet
[7:25] <iip_i2p> <Newsbyte> hi toad
[7:25] * tyesinclair (~tyesincla@170.252.64.1) has joined #freenet
[7:33] * Ash-Fox (Hal-9000@aas71.neoplus.adsl.tpnet.pl) Quit (Nick collision from services.)
[7:33] * Ash-Foxeus (Hal-9000@aak47.neoplus.adsl.tpnet.pl) has joined #FreeNET
[7:37] * mikeDOTd (~mikedotd@plutonium.nexthop.net) Quit (Read error: 54 (Connection reset by peer))
[8:13] <hobx_> bloog
[8:14] <silent-jaymz> goolb
[8:16] <hobx_> flagythonetic
[8:17] * |ux (~|ux@82-37-17-24.cable.ubr03.wolv.blueyonder.co.uk) has joined #freenet
[8:17] * greycat (~wooledg@192.35.79.70) has joined #freenet
[8:20] <|ux> hey greycat :)
[8:20] <greycat> hi
[8:20] <|ux> found out how to set the default OCM screen to connections
[8:20] <|ux> rather than peers
[8:21] <greycat> defaultToOCMHTMLPeerHandlerMode=false
[8:21] <greycat> right?
[8:21] <|ux> yep
[8:21] <|ux> i remember having a discussion the other day on here with someone saying that it should default to that... lol, never knew it was in the config file already
[8:22] <greycat> it's not in mine (by default)... but I haven't regenned my configs in quite some time. Someone on IIP/I2P/whatevertheycallitthismonth told me that magic string.
[8:23] <|ux> lol I was looking through the freenet.conf and I found it
[8:23] <|ux> along with a note on something that said "Opera and IE" are not properly secure and do not support proper anonymity via freenet?
[8:23] <greycat> IE is well known to be defective
[8:26] <|ux> how do i get freenet source?
[8:26] <greycat> CVS, or tarballs on snapshots
[8:27] <|ux> cvs from where?
[8:28] <greycat> :pserver:anonymous@cvs.freenet.sourceforge.net:/cvsroot/freenet
[8:28] <greycat> or at least that's what my CVS/Root file contains...
[8:29] <|ux> ok
[8:29] <|ux> is it like
[8:29] <|ux> CVSROOT=":pserver:anonymous@cvs.freenet.sourceforge.net:/cvsroot/freenet"
[8:29] <greycat> surely this is somewhere on the web pages...?
[8:29] <|ux> cvs checkout freenet
[8:31] * Zorix (Brandon@fl-64-45-226-120.dyn.sprint-hsd.net) Quit (Read error: 110 (Connection timed out))
[8:31] <|ux> ahh found it
[8:35] <hobx_> I think knowing how to use cvs is a proper hurdle to getting access to it.
[8:36] <|ux> hobx_: used to use it all the time, lol.. its easy to pick up again tho
[8:39] <hobx_> maybe freenet should move to subversion
[8:40] * bluephile (bluephile@69-160-205-193.clvdoh.adelphia.net) Quit (Client Quit)
[8:41] <|ux> ive only ever used cvs
[8:41] <|ux> , as long as it dont move to bitkeeper we are ok :p
[8:41] <|ux> oh wait, had to use this horrific CCC Harvest one at work before, ewwwwww
[8:43] <hobx_> Yeah. I don't know why Linus is such an idiot about the bitkeeper thing.
[8:54] <|ux> how many freenet nodes are there? rougly in total
[8:54] <|ux> roughly*
[8:54] <|ux> ie when everyone moves to 5089/5090 and so on.. how many nodes should there be on the network?
[9:16] * |ux (~|ux@82-37-17-24.cable.ubr03.wolv.blueyonder.co.uk) has left #freenet
[9:23] * ShaunMacPherson (~ShaunMac@Sudbury-HSE-ppp3980474.sympatico.ca) has joined #freenet
[9:33] * bluephile (bluephile@69-160-205-193.clvdoh.adelphia.net) has joined #freenet
[9:35] * |ux (~|ux@82-37-17-24.cable.ubr03.wolv.blueyonder.co.uk) has joined #freenet
[9:47] * Zorix (Brandon@fl-64-45-226-120.dyn.sprint-hsd.net) has joined #freenet
[9:48] <Redb3ard> dumb question guys, kinda offtopic
[9:48] <|ux> ??
[9:48] * |ux checks chanlog
[9:49] <Redb3ard> if you guys belonged to an invitation-only network, and you needed to invite, say 6 people, not from your own country, that would actually be interested...
[9:49] <Redb3ard> how difficult a task do you think that would be?
[9:49] <|ux> um impossible unless they see some purpose to it all
[9:49] <Redb3ard> assume the purpose is there
[9:50] <|ux> well, would they have the equipment / hardware/bandwidth / tech knowledge needed and time?
[9:50] <Redb3ard> not including myself, the best anyone has managed to do in a 6 month timeframe, is 2 invitations
[9:50] <|ux> ie how hard to set up, maintain,,, and how much of their computing resources woud it take
[9:51] <Redb3ard> those are all good questions, but the thing im trying to get at, i guess...
[9:51] <|ux> "Networking"?
[9:51] <Redb3ard> is how many people they actually talk to, and know well enough, online, that are even potential invitees
[9:51] <|ux> yea
[9:52] <|ux> "The Sociability factor" :p
[9:52] <Redb3ard> it just seems strange, that even online, people tend to cluster around a national boundary somehow
[9:53] <|ux> most people i know are from US
[9:53] <|ux> and im from UK :p
[9:53] <|ux> i guess it will have to do with timezones somewhat
[9:53] <|ux> for instance, when ur at home, they could be at work so u dont get a chance to talk
[9:55] <iip_i2p> <Newsbyte> I don't tend to cluster around a national boundary, I just don't like the USA ;-)
[9:55] <|ux> hehe
[9:55] <Redb3ard> hah
[9:56] <|ux> i think that the fact of how easy it is to install and leave alone would great affect it
[9:56] <iip_i2p> <Newsbyte> don't worry; I don't see all americans as idiotics ninkenpoops. There are individuals that are ok.
[9:56] <iip_i2p> <Newsbyte> don't worry; I don't see all americans as idiotics ninkenpoops. There are individuals that are ok.
[9:56] <|ux> i mean if htey can just download a 20kb file and run it... they wont mind really
[9:56] <iip_i2p> <Newsbyte> here and there
[9:56] <|ux> as long as they can get on with their browsing and IMing
[9:56] <iip_i2p> <Newsbyte> sometimes
[9:57] <iip_i2p> <Newsbyte> like:
[9:57] <|ux> newsbyte: lol
[9:57] <iip_i2p> <Newsbyte> ermm...ermm...
[9:57] <|ux> newsbyte: like those who come from another country?
[9:57] <Redb3ard> openvpn is the simplest setup
[9:57] <Redb3ard> but thats 1.5megs
[9:57] <iip_i2p> <Newsbyte> ;-)
[9:57] <Redb3ard> but it very nearly is just download it, and get on with browsing, IM and what not
[9:58] <|ux> is that like Bill gates saying "windows is a fully functioning operating system" then?
[9:58] <|ux> :p
[9:58] <iip_i2p> <Newsbyte> no wonder sanity moved from the US to europe :-)
[9:59] <|ux> lol
[9:59] <|ux> mv /dev/self /dev/country/urandom :p
[9:59] <Redb3ard> europe sane?
[9:59] <Redb3ard> lets not get in nationalistic pissing matches, please
[9:59] <|ux> redb3ard: sanity was a handle
[10:00] <Redb3ard> oh
[10:00] <Redb3ard> well, that makes much more sense, then
[10:00] <greycat> IRC user sanity == Ian Clarke
[10:01] <iip_i2p> <Newsbyte> hehe...yeah, an old inside joke, with that nick :-) I take it you aren't around long, redbeard?
[10:01] <Redb3ard> just been here a few days, really
[10:03] <iip_i2p> <Newsbyte> what build are you using?$
[10:03] * toad_ (toad@82-32-16-91.cable.ubr03.azte.blueyonder.co.uk) has joined #freenet
[10:03] <iip_i2p> <Newsbyte> ah, there, tha toad
[10:03] <iip_i2p> <Newsbyte> ah, there, tha toad
[10:03] * orasis (~orasis@c-66-41-30-194.mn.client2.attbi.com) has joined #freenet
[10:03] <iip_i2p> <Newsbyte> did you receive my latest mail, toadycody?
[10:04] <|ux> hey toad :)
[10:04] <iip_i2p> <Newsbyte> I mean, priv mail, not the maillist
[10:05] <iip_i2p> <Newsbyte> /me prods toad
[10:05] <toad_> hi
[10:05] <iip_i2p> <Newsbyte> What should I use for the FreenetScip?
[10:06] * toad_ has not read email yet
[10:06] <iip_i2p> <Newsbyte> I know the jar and seednodes...but if the build is drastically changed, should I include the exe and such as well?
[10:06] <|ux> Newsbyte: freenet-ext?
[10:07] <iip_i2p> <Newsbyte> has that changed?
[10:07] * lostlogic (~lostlogic@node-402418b2.mdw.onnet.us.uu.net) has joined #freenet
[10:07] <|ux> heh lolo
[10:07] <|ux> Newsbyte: no, but it did b4
[10:08] <lostlogic> hi, bye... gotta get to work (**late**) talk to ya soon
[10:08] * lostlogic (~lostlogic@node-402418b2.mdw.onnet.us.uu.net) Quit (Client Quit)
[10:08] <iip_i2p> <Newsbyte> but since it's not compatible with the latest builds, I tought it was time to update it
[10:08] <|ux> lol
[10:08] <iip_i2p> <Newsbyte> well, before 84 or after, because my latest FreentSCIP was 84
[10:09] <|ux> Newsbyte: ok
[10:10] <iip_i2p> <Newsbyte> what about nodeconfig and the distrib dir?
[10:13] * orasis is now known as Orasis
[10:13] * thelema is back
[10:14] <iip_i2p> <Newsbyte> /me prods toad (or are you emailing me the responses?)
[10:14] <iip_i2p> <Newsbyte> toad! You there?! It's time to change the FreenetSCIP, otherwise, some ppl will continue to install and use that obsolete build
[10:15] <iip_i2p> <Newsbyte> which is bad for the network, and thus for my ego
[10:15] <iip_i2p> <Newsbyte> dunno if that last correlation is worth anything, though ;-)
[10:19] <iip_i2p> <Newsbyte> well, you can easily augemtn it a bit, toad, say, 15% and 320MB
[10:27] <toad_> Newsbyte: what build was it last?
[10:27] <toad_> if it was 5084, you need to include the new freenet-ext.jar
[10:28] <toad_> but generally, include anything that has changed
[10:28] <iip_i2p> <Newsbyte> 84 (as I said)
[10:28] <iip_i2p> <Newsbyte> heh, yeah: what was changed? ;-)
[10:28] <iip_i2p> <Newsbyte> does it include the exe itself?
[10:31] <iip_i2p> <Newsbyte> I could update everything, fcourse, but that's a whole lot of d/l :-/ Especially when you have to search for it in snapshots. Maybe it could be presented a bit more structured, so unstable and stable thingies aren't all mixed?
[10:33] <toad_> freenet-ext.jar was definitely changed
[10:33] <toad_> i have no idea re the exe
[10:36] <iip_i2p> <Newsbyte> well...i'll start with the jar, ext-jar and seednodes, and see if I get complaints it ain't working ;-)
[10:36] <|ux> Newsbyte: freenet.conf ?
[10:49] <iip_i2p> <Newsbyte> did I miss anything?
[10:50] <iip_i2p> <Newsbyte> seednodes.ref sure is a huge package...
[10:50] * lolo-laptop (~lostlogic@68.251.84.226) has joined #freenet
[10:50] <|ux> wb lolo :)]
[10:50] <|ux> Newsbyte: why not use a smaller one?
[10:50] <lolo-laptop> :)
[10:50] <|ux> Newsbyte: http://82.37.17.24/buttons/seednodes.ref.gz
[10:50] * sanity (~ian@81-178-89-164.dsl.pipex.com) Quit (Read error: 104 (Connection reset by peer))
[10:50] <|ux> Newsbyte: works great for me (gunzip it when u get it)
[10:51] <iip_i2p> <Newsbyte> because I didn't find a smaller one on snapshots :-)
[10:51] * sanity (~ian@81-178-89-164.dsl.pipex.com) has joined #freenet
[10:51] * ChanServ sets mode +o sanity
[10:51] <|ux> Newsbyte: try that one ;) its 773k
[10:51] <iip_i2p> <Newsbyte> thnks, but it's almost d/l now
[10:51] <|ux> Newsbyte: remember the WinXP prob with "outOfMemory"
[10:52] <|ux> Newsbyte:im sure most ppl would prefer a smaller seednodes file
[10:54] * |ux thinks most people are himself :p
[10:54] <|ux> (joke)
[10:55] <hobx_> The funny thing is that the announcement protocol was designed to work best with as few seednodes as possible.
[10:55] <toad_> hobx_: no point doing that
[10:56] <|ux> hobx_: i thought that....
[10:56] <toad_> if we have 5 seednodes for the whole network then 99% of newbie nodes will take several days to get ONE connection
[10:56] <hobx_> Actually, there are several points. Even if you limit announcing to a small number of nodes in the file.
[10:56] <toad_> in any case it is quite possible to reduce the memory usage
[10:56] <toad_> hobx_: sure, in theory
[10:56] <hobx_> toad_: Well, you wouldn't seed everybdoy with the _same_ 5 nodes...
[10:56] <toad_> in theory, there are no load balancing issues. in theory freenet is easy!
[10:56] <iip_i2p> <Newsbyte> all good and well, but paranoids will want me to stay as close to the original as possible, if I accept seednodeslists from others...
[10:57] <toad_> hobx_: well we don't announce to all 100+ seednodes at once
[10:57] <hobx_> Actually, I think there are tons of load balancing issues in theory.
[10:57] <toad_> and even if you don't give everyone the same 5 nodes, you STILL need a huge list somewhere
[10:57] <toad_> also, 5 connections is SIMPLY NOT GOING TO WORK
[10:57] <greycat> you could make snapshots/seednodes.ref a CGI script that returns 20 random seeds
[10:58] <iip_i2p> <Newsbyte> yesyes, but do you have an offcicial, smaller seednodesfile, toad?
[10:58] <|ux> Newsbyte: the one i have i got from the 5090 build from that other place b4 it was upped to freenetproject.org
[10:59] <hobx_> The point is, you fill the routing table with lots of nodes from the seeds, then that undermines the discovered nodes of the announcement. This creates the broken topology we had before announcement where the majority of all queries would travel backwards in age.
[10:59] <iip_i2p> <Newsbyte> yes, ok...but who are you, ux? ;-)
[10:59] <|ux> Newsbyte: I see your point. I could be anyone
[10:59] <|ux> Newsbyte: and its |ux :p
[11:00] <hobx_> But my hands are washed, so...
[11:00] * jay (jcl@ool-18bf6dac.dyn.optonline.net) has joined #freenet
[11:00] <hobx_> bbl
[11:01] <toad_> |This creates the broken topology we had before announcement where the majority of all queries would travel backwards in age. - what on earth does that mean?
[11:01] * sanity (~ian@81-178-89-164.dsl.pipex.com) Quit (Read error: 104 (Connection reset by peer))
[11:01] <iip_i2p> <Newsbyte> oh, me, I'm all naive and gullible and all that. I belived Freenet was going to work within 5, years, for instance ;-) But paranoid users could rightfully point out I accept seednodes.ref from perfect strangers, who, maybe are placing sniffing seednodes or something
[11:01] <iip_i2p> <Newsbyte> you know how paranoid and crazy some freenetters are...even sanity seems sane, in comparison ;-)
[11:02] * interrupt (~chatzilla@64.122.23.213) has joined #freenet
[11:02] * pupok (~janie@81-178-89-164.dsl.pipex.com) Quit (Read error: 104 (Connection reset by peer))
[11:03] * pupok (~janie@81-178-89-164.dsl.pipex.com) has joined #freenet
[11:03] <iip_i2p> <Newsbyte> so, toad, any chance on pointing me to a lighter, yet official ref?
[11:03] <|ux> Newsbyte: i get the point....
[11:03] * pupok (~janie@81-178-89-164.dsl.pipex.com) Quit (Read error: 104 (Connection reset by peer))
[11:03] <iip_i2p> <Newsbyte> no point in creating frustration with XP users...
[11:04] * pupok (~janie@81-178-89-164.dsl.pipex.com) has joined #freenet
[11:04] <iip_i2p> <Newsbyte> you get the point of me being gullible and all that too? ;-)
[11:04] * pupok (~janie@81-178-89-164.dsl.pipex.com) Quit (Read error: 104 (Connection reset by peer))
[11:04] * silent-jaymz (~jaymz@dialup-226.82.221.203.acc52-dryb-mel.comindico.com.au) Quit ("leaving")
[11:04] * jokern (~virus@213.184.217.12) has joined #freenet
[11:04] <|ux> Newsbyte: hehe. I get the point you are sensible enough to consider others.
[11:05] * pupok (~janie@81-178-89-164.dsl.pipex.com) has joined #freenet
[11:05] <iip_i2p> <Newsbyte> blasfemy!! Me, sensible! phooh!
[11:05] <iip_i2p> <Newsbyte> toad can assure you that I'm not! Isn't that right, toad?
[11:07] <|ux> ahh but my seednodes.ref file has all .gov ip addresses in it (joke)
[11:07] <iip_i2p> <Newsbyte> he is agreeing in silence :-)
[11:07] <|ux> toad_: finally found out how to default the fproxy screen to "connections mode" rather than peers mode for OCM
[11:08] <jokern> Is toad_ here?
[11:08] <toad_> yes
[11:09] <KenMan> i have a picture for the amphibian ... but would appreciate input from anyone regarding comprehensibility : http://mywebpages.comcast.net/jkcorson/FreeNet/backoff.png
[11:09] <jokern> I am still loosing alot of keys when restarting node
[11:09] <|ux> toad_; it was in the config all the time
[11:09] * Kyon (~Kyon@pcd396122.netvigator.com) has joined #freenet
[11:09] <toad_> KenMan: there are so many things in your mail that i don't understand or i don't agree with... read the reply...
[11:10] <KenMan> well, thank you for the reply(s), anyway :)
[11:10] <iip_i2p> <Newsbyte> me would still like to know what ref to take...
[11:10] <toad_> KenMan: nice graphic
[11:10] <toad_> we should log % backed off as well as backedOffCount... maybe sample it every 30 secs or something...
[11:11] <|ux> http://freenetproject.org/snapshots/seednodes.ref.bz2
[11:11] <KenMan> I am already doing that, remember ??
[11:11] <toad_> good
[11:11] <jokern> toad_ : Before restart-> Total keys: 8489 After restart -> Total keys: 6573
[11:11] <toad_> maybe it should go on the general stats, i mean
[11:11] * sanity (~ian@81-178-89-164.dsl.pipex.com) has joined #freenet
[11:11] * ChanServ sets mode +o sanity
[11:11] <toad_> anyway your point is that very low average backedOffCount does not mean we don't have very high overall backoff %
[11:11] <toad_> right?
[11:11] <toad_> hi sanity
[11:12] <KenMan> yes . i achieved communication for once :)
[11:12] <toad_> KenMan: I really don't see why there is any benefit to sending the same MRI out to all requestors
[11:12] * sanity (~ian@81-178-89-164.dsl.pipex.com) Quit (Read error: 54 (Connection reset by peer))
[11:12] * pupok (~janie@81-178-89-164.dsl.pipex.com) Quit (Read error: 104 (Connection reset by peer))
[11:12] <toad_> ugh
[11:12] <KenMan> uh oh, we scared him off :o
[11:12] <jay> is he on mirc ?
[11:12] <toad_> yeah, he might have to do some work :)
[11:13] * sanity (~ian@81-178-89-164.dsl.pipex.com) has joined #freenet
[11:13] * ChanServ sets mode +o sanity
[11:13] <jay> heh
[11:13] <toad_> hi sanity again :)
[11:13] * pupok (~janie@81-178-89-164.dsl.pipex.com) has joined #freenet
[11:13] <jay> ah a Mac
[11:13] <toad_> KenMan: okay, convince me: why do we need to give equal quota to each node?
[11:13] <jay> w/ xchat.. nice
[11:13] <sanity> weird, my connection keeps dropping for soe reason
[11:14] <KenMan> it would improve the current situation, is all. We don't strictly have to.
[11:14] <toad_> sanity: KenMan's had some Deep Thoughts on rate limiting
[11:14] <toad_> KenMan: how would it improve anything ?
[11:14] <|ux> toad_: seednodes.ref ? use the Massive one or not?
[11:14] <toad_> the only reason there is imbalance is because nodes don't use the quota we give them
[11:14] <KenMan> I'm out of arguing energy right now (and lacking some sleep)...
[11:15] <KenMan> ah hah, you are onto part of the problem :)
[11:15] * sanity (~ian@81-178-89-164.dsl.pipex.com) Quit (Read error: 104 (Connection reset by peer))
[11:15] * pupok (~janie@81-178-89-164.dsl.pipex.com) Quit (Read error: 104 (Connection reset by peer))
[11:15] <jay> whos pupok?
[11:15] <jay> ian's clone?
[11:15] <KenMan> ian's unmarried wife ?
[11:15] <toad_> KenMan: I fail to see how nodes not using the quota we give them could possibly be caused by being given high MRIs
[11:15] * pupok (~janie@81-178-89-164.dsl.pipex.com) has joined #freenet
[11:15] <toad_> ian's gf, yes
[11:16] <jay> ian's planning a takeover of #freenet
[11:16] * pupok (~janie@81-178-89-164.dsl.pipex.com) Quit (Read error: 104 (Connection reset by peer))
[11:16] <jay> join/leave flood
[11:17] * pupok (~janie@81-178-89-164.dsl.pipex.com) has joined #freenet
[11:17] <iip_i2p> <Newsbyte> seednodes.ref, smaller?
[11:17] <KenMan> toad - did you get my demo ? and did it mean anything to you ?
[11:17] <jay> KenMan/Toad are have a different discussion than the one |ux/Newsbyte are having
[11:18] <toad_> it didn't mean much to me
[11:18] <KenMan> the fact that 50% of queries can't get past my node doesn't shock you slightly ? or was my patch completely incorrect ?
[11:20] <KenMan> the fact that 50% of queries run into complete and total backoff (thus RNF) doesn't surprise you ?? This is likely a common thing on most nodes.
[11:20] <KenMan> As in "FreeNet uses a broadcast topology" ??
[11:21] <toad_> KenMan: sure
[11:22] <toad_> but I fail to see the relationship between that and equal quota for nodes that DONT USE THEIR EXISTING QUOTA
[11:22] * theLostFloppy (~michaelku@fia41-111.dsl.hccnet.nl) has joined #freenet
[11:22] <|ux> anyone give reasons why the config file says Opera and IE are insecure for anonymity in freenet?
[11:22] <jay> |ux: it's been discussed on the Dev list
[11:22] <toad_> |ux: because IE doesn't respect MIME types
[11:22] <toad_> ever
[11:23] <hobx_> toad_: When seeding was the only form of node introduction, you end up with a case where new nodes learn about older nodes, but never vice versa.
[11:23] <Kyon> |ux: doesnt support MIMEs properly, i think
[11:23] * pupok (~janie@81-178-89-164.dsl.pipex.com) Quit (Read error: 104 (Connection reset by peer))
[11:23] <toad_> and Opera doesn't unless you tell it to
[11:23] <KenMan> forget about quota until you see the need for it... something like 20-60% probability of RNF at EACH HOP is kind of a bad thing for routing , no ? How deep can most queries go in a situation like that ??
[11:23] <toad_> KenMan: sure
[11:23] <|ux> kk thx guys :)
[11:23] <Kyon> |ux: so IE might open a .html which is really plain text as an html file
[11:23] <toad_> the estimators say 70% probability of RNF
[11:23] <|ux> lol
[11:23] <toad_> at EACH HOP
[11:23] <toad_> which is rather bad
[11:23] * pupok (~janie@81-178-89-164.dsl.pipex.com) has joined #freenet
[11:23] <hobx_> toad_: When seeding is the primary form of node introduction (because announcement doesn't work well, and you compensate with many seeds) you get mostly the same situation.
[11:24] <KenMan> doesn't it motivate you to want to abolish retries and let the network work ?
[11:24] <Kyon> |ux: btw - you probably dont remember, but i disconnected in the middle of talking to you yesterday. I had a power outage, wanted to apologize ;)
[11:24] <toad_> hobx_: I'm not sure that's the main reason
[11:24] <toad_> KenMan: huh?
[11:24] * pupok (~janie@81-178-89-164.dsl.pipex.com) Quit (Read error: 104 (Connection reset by peer))
[11:24] <toad_> KenMan: hmmm
[11:24] <KenMan> toad: huh ?
[11:24] <KenMan> toad: yes, hmmm...
[11:24] <toad_> what makes you think that retries on RNF are a problem?
[11:24] <hobx_> I don't know if it is the main reason (for what?) but it is certainly a reason why too much seeding is a very bad thing.
[11:24] <toad_> they are limited by HTL
[11:25] * toad_ is ignoring hobx_ for now
[11:25] * pupok (~janie@81-178-89-164.dsl.pipex.com) has joined #freenet
[11:25] <KenMan> you still don't get it. Damn !!! How to explain this ?? I'll walk away for some minutes, and come back.
[11:25] <toad_> KenMan: I don't see how RNFs can result in request multiplication across the network
[11:25] <hobx_> You asked, suit yourself.
[11:25] <toad_> on one node, perhaps
[11:25] <hobx_> Plus, I'm on topic for once...
[11:25] * sanity (~ian@81-178-89-164.dsl.pipex.com) has joined #freenet
[11:25] * ChanServ sets mode +o sanity
[11:26] <toad_> but generally, the retries caused by RNFs will be compensated for by requests expiring when they run out of HTL
[11:26] <toad_> so where does this happen?
[11:26] <|ux> Cant users be given a choice?
[11:26] <toad_> |ux: about what?
[11:26] <|ux> "Offical Massive Seednode File.. or Unofficial Seednode File but smaller"
[11:26] * sanity (~ian@81-178-89-164.dsl.pipex.com) Quit (Read error: 104 (Connection reset by peer))
[11:27] <toad_> sure, there are several smaller ones
[11:27] <|ux> a bit like .met list for eMule
[11:27] <toad_> the reason the seednodes file is huge is, mainly, that if it wasn't, all the seednodes would be overloaded
[11:27] <toad_> and I don't see any good reason to cut it down
[11:27] <toad_> if we wanted to cut it down we'd just choose fewer nodes when reseeding
[11:27] <|ux> So that WinXP bug "OutOfMemory" thing is?
[11:27] * pupok (~janie@81-178-89-164.dsl.pipex.com) Quit (Read error: 104 (Connection reset by peer))
[11:27] <toad_> but I see absolutely no reason not to fill the RT up with seednodes
[11:28] <toad_> |ux: I suppose I ought to fix that
[11:28] * pupok (~janie@81-178-89-164.dsl.pipex.com) has joined #freenet
[11:28] <toad_> or we could just increase the -Xmx again :)
[11:28] <dERMOTH> damn my system is already out of memory, even the web browsed not running :(
[11:28] * moskau23 (~Miranda@dsl-082-082-224-034.arcor-ip.net) Quit (Read error: 104 (Connection reset by peer))
[11:29] <dERMOTH> why not a cgi script?
[11:29] <toad_> dERMOTH: why not do it on the client end?
[11:29] <dERMOTH> whi not?
[11:29] <dERMOTH> why
[11:29] <greycat> toad_: because downloading 20 MB seednodes files is bad?
[11:30] <toad_> greycat: except that you're actually downloading 900kB
[11:30] <dERMOTH> it should also download the gziped version
[11:30] <toad_> dERMOTH: huh? it doesn't?
[11:30] <dERMOTH> no
[11:30] <toad_> well that's just the wininstaller crap for you
[11:30] <toad_> update.sh does
[11:30] <dERMOTH> maybe i got an old update.gz?
[11:31] <dERMOTH> wget http://freenetproject.org/snapshots/seednodes.ref -O new-seednodes.ref
[11:31] <toad_> dERMOTH: old update.sh
[11:31] <toad_> iirc
[11:32] <dERMOTH> well, crap! I'm gonna rm -rf freenet!!
[11:32] <dERMOTH> and reinstall it all over again
[11:33] <toad_> the String form of a reference is considerably smaller than the FieldSet version, right?
[11:33] <dERMOTH> too bad for the store and routing table
[11:35] <toad_> bbiab, going to eat
[11:35] * pupok (~janie@81-178-89-164.dsl.pipex.com) Quit (Read error: 104 (Connection reset by peer))
[11:35] <toad_> then i may be back briefly
[11:36] <toad_> then i'll probably go to aikido around 18:20
[11:36] <iip_i2p> <Newsbyte> so...what (smaller, official) seednodes.ref can I use, now?
[11:36] <toad_> and come back around 22:00
[11:36] * pupok (~janie@81-178-89-164.dsl.pipex.com) has joined #freenet
[11:36] <Kyon> are the unstable/stable nets shared at all? if not, which one is more active, stable?
[11:36] * hobx_ wonders if toad takes off his backpack during aikido...
[11:36] * pupok (~janie@81-178-89-164.dsl.pipex.com) Quit (Read error: 104 (Connection reset by peer))
[11:37] * pupok (~janie@81-178-89-164.dsl.pipex.com) has joined #freenet
[11:37] * sanity (~ian@81-178-89-164.dsl.pipex.com) has joined #freenet
[11:37] * ChanServ sets mode +o sanity
[11:38] <hobx_> I think a certain couple should fix their internet connection...
[11:38] * sanity (~ian@81-178-89-164.dsl.pipex.com) Quit (Read error: 104 (Connection reset by peer))
[11:38] * pupok (~janie@81-178-89-164.dsl.pipex.com) Quit (Read error: 104 (Connection reset by peer))
[11:38] * pupok (~janie@81-178-89-164.dsl.pipex.com) has joined #freenet
[11:38] <iip_i2p> <Newsbyte> I think you might be right
[11:40] * pupok (~janie@81-178-89-164.dsl.pipex.com) Quit (Read error: 104 (Connection reset by peer))
[11:40] <|ux> lol
[11:40] <KenMan> toad: read last night's backlog where I proceed to trash you.
[11:40] <KenMan> Then consider this philosophical question:
[11:40] * pupok (~janie@81-178-89-164.dsl.pipex.com) has joined #freenet
[11:40] <iip_i2p> <Newsbyte> well, stable is bigger, so I guess that's more active (in the sense of content), but in comparison, unstable is more active, relative percentage, me thinks
[11:41] <KenMan> and let's talk about nodes giving each other quota in terms of qph, rather than using MRIs directly.
[11:41] <iip_i2p> <Newsbyte> oh-oh...philosophical questions suck!
[11:41] <iip_i2p> <Newsbyte> ;-)
[11:41] <greycat> I'd rather talk about getting my node to not have 80 active transfers
[11:41] * |ux redirects KenMan to freenet-chat
[11:41] <|ux> lol
[11:41] <KenMan> Newsbyte - shut up for 3 minutes. I'm trying to fix freenet.
[11:41] * pupok (~janie@81-178-89-164.dsl.pipex.com) Quit (Read error: 104 (Connection reset by peer))
[11:42] <Kyon> newsbyte?
[11:42] * pupok (~janie@81-178-89-164.dsl.pipex.com) has joined #freenet
[11:42] * pupok (~janie@81-178-89-164.dsl.pipex.com) Quit (Read error: 104 (Connection reset by peer))
[11:42] <hobx_> think Ian will be pissed off if I ban his subnet?
[11:43] <Kyon> nm
[11:43] <KenMan> suppose we have a network of 12723 nodes. We stop time and freeze the network. Now, if we add the quota (in qph) of each node that is outgoing, and we add the quota (in qph) of each node this is incoming, can they be different ?
[11:43] <greycat> hobx_: probably
[11:43] <KenMan> i state that they cannot. As each node has to receive the quota sent by another node.
[11:44] <KenMan> this is the foundation for me saying that in quota must be approximately equal to out quota.
[11:44] <dERMOTH> What'S the recommended j2se version? 1.4.1, 1.4.2 or 1.5 beta??
[11:44] <hobx_> KenMan: They can't. The number of extra outgoing (those generated by the nodes themselves) have to match extra incoming (queries terminating).
[11:44] <hobx_> It is pure logic.
[11:45] <KenMan> We can't truly multiply our queries on a collective basis. Route exhaustion leading to local RNFs are the hidden adjustment factor.
[11:45] <dERMOTH> any1 have a clue?
[11:46] <hobx_> dERMOTH: I would avoid the beta
[11:46] <KenMan> for every retry sent out (in response to getting an RNF back), another request gets blocked locally. And loses its chance to move further into the network.
[11:47] <KenMan> so the average depth of all queries is somewhere close to 3 sequential hops. Because the HTL gets used up in retries.
[11:47] <KenMan> am I making sense yet ? I can keep yammering, if not.
[11:48] <dERMOTH> if i remember well, long time ago i heard of some problems with newer version of java...So I still use 1.4.1... Should I upgrade?
[11:48] <hobx_> KenMan: Not to me, but maybe to somebody else.
[11:49] <KenMan> dERMOTH: I have wonderful results with 1.4.2_03 ...
[11:49] <iip_i2p> <Newsbyte> you mean 6
[11:49] <iip_i2p> <Newsbyte> der, no and yes
[11:49] <iip_i2p> <Newsbyte> der, no and yes
[11:49] <iip_i2p> <Newsbyte> der, no and yes
[11:49] * KenMan doesn't upgrade regularly
[11:49] <iip_i2p> <Newsbyte> the eternal diplomatic answer ;-)
[11:49] * sanity (~ian@81-178-89-164.dsl.pipex.com) has joined #freenet
[11:49] * ChanServ sets mode +o sanity
[11:49] * sanity (~ian@81-178-89-164.dsl.pipex.com) Quit (Client Quit)
[11:50] <iip_i2p> <Newsbyte> nonsense! 1.4.2 is, genrally speaking, no good
[11:50] <jay> dERMOTH: the latest 1.4.2 is 1.4.2_05
[11:50] <jay> dERMOTH: and it works well for me, even with -server
[11:50] <iip_i2p> <Newsbyte> or you stay with 1.4.1, or you use 1.5
[11:50] <iip_i2p> <Newsbyte> or you stay with 1.4.1, or you use 1.5
[11:50] <KenMan> hobx_: if every request stands only a 50% chance of making it through a node, how far can it get with HTL=20 ? generally ?
[11:51] <jay> Newsbyte: i wouldn't give that advice
[11:51] <jay> dERMOTH: what os?
[11:51] <hobx_> KenMan: You mean a 50% chance of being rejected?
[11:51] <dERMOTH> redhat 7.3
[11:51] <KenMan> hobx: yes. the problem is that a significant percent of requests are experiencing route exhaustion - due to backoff.
[11:51] <iip_i2p> <Newsbyte> why the f- does everything I type comes out double or triple, suddenly?
[11:52] <jay> dERMOTH: linux on i86? id try 1.4.2_05 first
[11:52] <iip_i2p> <Newsbyte> that could be that you wouldn't, but I am: 1.4.2 sucks, most agree on that
[11:52] <hobx_> Well, if 50% are rejected, then each node will on average get one rejection before finding a node that accepts.
[11:52] <dERMOTH> k
[11:52] <KenMan> hobx - thank you for restating the problem for me.
[11:53] <iip_i2p> <Newsbyte> look, jay, there are a myriad of complaints when using 1.4.2, it is NOT recommended, even by Toad himself
[11:53] <hobx_> So only 10 of 20 hops would be forward
[11:53] <jay> Newsbyte: do you run windows?
[11:53] <jay> dERMOTH runs linux.. the latest 1.4.2 doesn't show the same problems that *older* 1.4.2 vm's did
[11:53] <KenMan> now, suppose that the other 50% never get an opportunity to issue a request beyond the local node.
[11:53] <jay> i verified it myself on linux
[11:53] <hobx_> but IIRC even rejecting nodes do check if they have the data, so that helps somewhat.
[11:54] <hobx_> KenMan: I'm not sure what that means.
[11:54] <jay> dERMOTH: i wouldn't delete the old JSDK.. install them side by side
[11:54] <hobx_> What was the problem with 1.4.2?
[11:55] <jay> hobx_: it randomly sig 11'd on my box
[11:55] <jay> 1.4.2_03
[11:55] <|ux> i use 1.4.2
[11:55] <iip_i2p> <Newsbyte> yes, yes, well I still don't recommend it. If 1.4.1 works, don't try to fix it. If you must update, don't do it with 1.4.2, that's what I recommend. Feel free to neglect my advice, of course.
[11:55] <|ux> it works fine for me
[11:55] <hobx_> bad RAM
[11:55] <jay> hobx_: negative
[11:55] <jay> |ux: what sub-version?
[11:55] <jay> each JDK version is 1.4.x_y
[11:55] <KenMan> hobx - when queries enter my node at a specific moment, there is a 50% likeliehood that ALL routes are currently unavailable. The only choice is to immediately send back an RNF.
[11:56] <jay> im saying 1.4.2_05 works fine
[11:56] <hobx_> KenMan: Ok.
[11:56] <|ux> 1.4.2_04
[11:56] * sanity (~ian@81-178-89-164.dsl.pipex.com) has joined #freenet
[11:56] * ChanServ sets mode +o sanity
[11:56] <KenMan> hobx - i believe this is fairly consistent on many if not all nodes
[11:56] <jay> |ux: okay i had the problems with 1.4.2_03
[11:57] <KenMan> hobx - I believe this is common on all nodes. Not just some of them. The 50% wanders between 30-70%, but it happens everywhere.
[11:57] <KenMan> bye sanity (might as well get it out of the way now)...
[11:57] * sanity (~ian@81-178-89-164.dsl.pipex.com) Quit (Read error: 104 (Connection reset by peer))
[11:57] <|ux> LOL
[11:57] <|ux> :P
[11:57] <iip_i2p> <Newsbyte> no, you're saying 1.4.2-05 works fine for you. I can't imagine you have a list of ppl who tried out 1.4.2-05 after the definately crappy 1.4.2 and decided all probs were solved, now?
[11:58] * sanity (~ian@81-178-89-164.dsl.pipex.com) has joined #freenet
[11:58] * ChanServ sets mode +o sanity
[11:58] <dERMOTH> I like the "more" disclaimer where you just have to press "q" then pretend you have read it all :)
[11:58] <hobx_> so basically what is happening is you used to have trouble with nodes rejecting because of overload, but now you back off those instead. But then you just moved the problem because all nodes are overloaded.
[11:58] * sanity (~ian@81-178-89-164.dsl.pipex.com) Quit (Read error: 104 (Connection reset by peer))
[11:58] <|ux> dERMOTH: i agree :p
[11:58] <jay> Newsbyte: im not going to play verbal chess with you.. what i said is what I've experienced and i said no more
[11:58] <hobx_> What I don't understand when I hear these things is I just talked to somebody who claimed that Freenet was working very well ATM.
[11:59] <jay> hobx_: compared to recently it is
[11:59] <|ux> it is working the best ive ever seen it :p
[11:59] <jay> Newsbyte: funny how it's working without the fixed-cluster size eh?
[12:00] <iip_i2p> <Newsbyte> good, and what I say is, that at least 4 ppl on the maillist and two o?n the wiki said they had probs with 1.4.2, and that 1.4.2 isn't recommended, even by toad himself.
[12:00] <jay> Newsbyte: yes... 1.4.2_03 !!!!!!!!!! 03 03 03 03
[12:00] <jay> read u troll
[12:00] <KenMan> hobx - the percent of routes unavailable is a direct function of how many times a random query gets retried. Thus exceeding that query's "philosophical" single quota slot. Ins=outs, remember ?
[12:00] <jay> nothing but compaints from u
[12:00] <|ux> jay: :p lol
[12:01] <KenMan> rather, the percent of time that a query hits a brick wall is a fn of the average multiplication factor for those queries that don't hit the brick wall.
[12:01] <jay> the only time Newsbyte has a discussion in here is to fire off commands and get everyone whipped up over a brand new feature
[12:01] <hobx_> KenMan: Ins = Outs globally. But a QR doesn't change that equation, since the sending node gets and extra out, and rejecting gets an extra in.
[12:01] <|ux> im out
[12:01] <|ux> be back tonight
[12:02] <|ux> later guys
[12:02] * |ux (~|ux@82-37-17-24.cable.ubr03.wolv.blueyonder.co.uk) Quit ("Leaving")
[12:02] <KenMan> hobx - in the brave new world of query limiting, we cannot account for anything 'extra' :(
[12:02] <jay> and toad *has* told me to upgrade to 1.5.0 but i haven't since my node is working fine
[12:02] <KenMan> a slot in time gets blacked out
[12:02] <greycat> When life gets you down, just recite the Freenet Haiku: Java's not so bad / My Freenet node seems stabl-- / OutOfMemory
[12:03] <KenMan> greycat: HAHAHA
[12:03] <KenMan> :)
[12:03] <Kyon> greycat: heh
[12:03] <iip_i2p> <Newsbyte> nonsense.they said 1.4.2 without specifying it, at least in most cases. whether or not subversions are better, is left to wonder, and unless there is a general consensus that a particular subversion is suddenly ok, it's prudent to not recommend it.
[12:03] <greycat> is blackdown 1.4.2-rc1 considered evil?
[12:04] <jay> Newsbyte: upon further discussion on *my* part in this channel i determined that older 1.4.2 sdk's were generating sig 11's on linux
[12:04] * moskau23 (~Miranda@dsl-213-023-250-039.arcor-ip.net) has joined #freenet
[12:04] <jay> Newsbyte: they didn't claim which 1.4.2 was causing them problems in the email
[12:04] <greycat> mine has crashed 2-3 times on me with segv
[12:04] <greycat> but it's pretty rare
[12:04] <jay> greycat: 1.4.2?
[12:04] <greycat> blackdown 1.4.2-rc1
[12:05] <jay> ah these people were using Sun's
[12:05] <iip_i2p> <Newsbyte> not true, again, I often discus in national pissing contests too. Mostly anti-usa and such :-)
[12:05] <dERMOTH> i don't even care it it crasn if it can help me on performances..
[12:05] <jay> Newsbyte: yes more arguing
[12:05] <dERMOTH> if it crash
[12:05] <greycat> jay: nah, that's just trolling
[12:06] <hobx_> KenMan: I have been around for five years, and every such policy has blown up and made matters worse. Every single.
[12:06] <iip_i2p> <Newsbyte> hmm...depends how you look at it, I guess
[12:06] <jay> Newsbyte: i come here to code and see Freenet in development
[12:06] <iip_i2p> <Newsbyte> I *did* make a post 'how not to feed the trolls' on /. ;-)
[12:06] <jay> and i get sucked into arguments
[12:07] <jay> sometimes
[12:07] <jay> good thing they don't hand out ops in here
[12:07] <iip_i2p> <Newsbyte> that said, ppl mostly say I'm a troll, because they find it annoying that I say it as it is, at least in my view
[12:07] <jay> all trolls are speaking from their own point of view
[12:07] <jay> that doesn't make the observation important
[12:07] <iip_i2p> <Newsbyte> and it's mostly the same ppl calling me a troll too
[12:07] <toad_> hi
[12:07] <KenMan> hobx - i am calmed by your comment. So the frustration i feel is just useless energy ... it cannot be applied. Gotcha ;)
[12:08] <jay> Newsbyte: im one of them i guess
[12:08] <hobx_> I must be the biggest troll, because I have had my IP permanently banned from slashdot...
[12:08] <iip_i2p> <Newsbyte> ah well...if I'm a troll, at least I'm a troll with cvs access ;-)
[12:08] <jay> hobx_: haha
[12:08] <hobx_> jay: Funny for you.
[12:08] <iip_i2p> <Newsbyte> indeed, and you are speaking from your view, when recommending 1.4.2-05
[12:08] <toad_> KenMan: here?
[12:08] <KenMan> i am here
[12:09] <jay> Newsbyte: you have cvs? i do as well so i can undo your trollish changes ;)
[12:09] <toad_> KenMan: okay
[12:09] <toad_> what do you want me to read?
[12:09] <jay> Newsbyte: the help stuff you've done is cool
[12:09] <KenMan> toad: do you mean it ? Do you really really mean it ??
[12:09] <toad_> you said giving it out in requests per hour might help
[12:09] <toad_> why would it be any different from what the current situ is
[12:09] <toad_> ?
[12:09] * sanity (~ian@81-178-89-164.dsl.pipex.com) has joined #freenet
[12:09] * sanity (~ian@81-178-89-164.dsl.pipex.com) Quit (Client Quit)
[12:09] <jay> Newsbyte: yes but i actually helped figure out specifics
[12:09] <KenMan> might help in understanding the problem, not changing it.
[12:10] <jay> Newsbyte: like which VM was causing the errors, since it worked for some and not for others
[12:10] <iip_i2p> <Newsbyte> I'm not speaking from my view when not recommending it, I'm speaking from the view of all those ppl that had probs with it. Whather or not all probs are solved would be premature to say, even considering your personal experiance, I think that speaks for itself
[12:10] <jay> Newsbyte: and as a programmer i like to know details before i give out advice
[12:10] <toad_> <KenMan> this is the foundation for me saying that in quota must be approximately equal to out quota. - yup
[12:10] <toad_> this is reasonable
[12:10] <toad_> so we need a way to make in = out, which doesn't lead to a network meltdown
[12:11] <jay> Newsbyte: if that person comes back and says "1.4.2_05 is crashing" ill change my thinking
[12:11] <KenMan> toad, if you disable retries on RNF, the network will go to heaven.
[12:11] <toad_> and you're saying that the way for that to happen is to ensure, on every node, all the time, that incoming requests > outgoing requests
[12:11] <toad_> right?
[12:11] <toad_> KenMan: what do you mean by disable them?
[12:11] <KenMan> yeah man, you're in hannah now
[12:11] <iip_i2p> <Newsbyte> hah...a bit too late for HIM then, won't it? ;-)
[12:11] <jay> Newsbyte: do u post on Frost?
[12:11] <toad_> hannah?
[12:11] <jay> Newsbyte: 1 crash? no biggie
[12:11] <hobx_> There is a tendency to think that nodes where out > in are "bad guys" causing the problem, and in > out are helping. But in fact it is the opposite.
[12:11] <toad_> if we get a QueryRejected, return a QueryRejected to the previous node?
[12:11] <jay> Newsbyte: i would then send the mail to Dev about it
[12:11] <toad_> hobx_: eh?
[12:12] <hobx_> Nodes with in > out are just rejecting.
[12:12] <toad_> hobx_: out > in is bad
[12:12] <iip_i2p> <Newsbyte> that's not for you to decide. 1 crash for someone else can be sufficient to drop Freenet
[12:12] <KenMan> i mean, don't ever retry on an RNF. I think that 90%+ are caused by route exhaustion, and <10% are actual routing loops.
[12:12] <toad_> hobx_: well, we don't reject much because of rate limiting
[12:12] <hobx_> toad_: out > in is a working node trying to find a neighbor that works.
[12:12] <KenMan> Therefore, if applied net wide, the ONLY reason for RNFs would be looping. Well, approximately anyway.
[12:12] <hobx_> toad_: Accoding to Ken his node hovers between 30-70%
[12:12] <jay> Newsbyte: here you go again arguing.. you help your way and ill do the same..
[12:12] <toad_> hobx_: but if out > in, then things suck
[12:13] <toad_> hobx_: 30%-70% rejected?
[12:13] <hobx_> toad_: No, if things suck, then out > in.
[12:13] <toad_> you mean initially rejected, or QR'd?
[12:13] <hobx_> You have the causality backwards.
[12:13] <KenMan> you see, we are attempting to do out = 3*in , but backoff squelches it
[12:13] <toad_> what we are talking about here is late QRs
[12:13] <jay> Newsbyte: but gather facts and specifics, not general ideas before giving advice to new users
[12:13] <toad_> QR after Accepted
[12:13] <iip_i2p> <Newsbyte> it just doesn't make sense to recommend something where probs are being reported, just because some version works with you. the best advice is: if it's not broke, don't fix it
[12:13] <toad_> which means RouteNotFound
[12:13] * jay sighs
[12:13] <jay> whatever
[12:13] <toad_> KenMan: if we get a long RNF, we will significantly reduce the HTL, right now
[12:14] <iip_i2p> <Newsbyte> indeed! whatever! as long as it's not 1.4.2! ;-)
[12:14] <toad_> hobx_: if every node outputs 100% more queries than it receives, what do you think will happen to the network?
[12:14] <KenMan> and thus, the request cannot go very deep. Now imagine this happening to every other node.
[12:14] <KenMan> rather, every other query
[12:14] <hobx_> Kenman: "hobx - when queries enter my node at a specific moment, there is a 50% likeliehood that ALL routes are currently unavailable. The only choice is to immediately send back an RNF."
[12:15] <toad_> KenMan: is that measured?
[12:15] <hobx_> toad_: But that can't happen. Nodes are outputing more than they get because other nodes are not doing any work.
[12:15] <KenMan> sadly, yes. My demo was actually meant to be used by you.
[12:15] <hobx_> Think about it.
[12:15] <toad_> hobx_: well that's an RNF
[12:15] <toad_> that's not an initial QueryRejected, it occurs after routing
[12:15] <toad_> hobx_: nodes are retrying after other nodes RNF them, right?
[12:15] <KenMan> which amounts to the same thing
[12:15] <hobx_> it is still an In with no Out.
[12:16] <KenMan> exactly
[12:16] <toad_> hobx_: so one way to solve it would be to not retry on QueryRejected?
[12:16] <KenMan> yes
[12:16] <toad_> and implement per-node failure tables
[12:16] <toad_> it
[12:16] <KenMan> unfortunately, that means loops will not get fair treatment.
[12:16] <toad_> it's a PITA re loops though...
[12:16] <hobx_> then all queries will just fail
[12:16] <KenMan> but i don't think they represent anywhere near the majority of RNFs
[12:16] <toad_> hobx_: will they?
[12:17] <hobx_> unless I misunderstand you
[12:17] <toad_> if all nodes fail requests on an RNF, then why would all queries just fail?
[12:17] <toad_> obviously if we just set that on ONE node, we'd be in trouble
[12:17] <KenMan> actually you might see higher success
[12:17] <toad_> KenMan: perhaps
[12:17] <hobx_> Well, you can try.
[12:17] <KenMan> but it needs to be netwide
[12:18] <toad_> we would get lower MRIs certainly
[12:18] <KenMan> toad - run my demo and crap your pants, please.
[12:18] <toad_> and less backoff
[12:18] <hobx_> It might help against overload at the cost of failing queries have just taken an unlucky turn.
[12:18] <toad_> KenMan: please give me the numerical results
[12:18] <toad_> hobx_: that already happens. DataNotFound.
[12:18] <toad_> combined with backoff
[12:18] <toad_> also the fact that we have to decrement htl on QueryRejected's
[12:18] <hobx_> well, I'm all for giving up earlier.
[12:18] <toad_> and significantly reduce it on RNFs
[12:19] <toad_> hobx_: you are?
[12:19] <hobx_> I have been trying to say for a while that one of the biggest problems is the network tries too hard.
[12:19] <toad_> so has thelema
[12:19] <KenMan> and now, me too .
[12:19] <hobx_> If don't try do hard: then overloaded network = degraded results
[12:20] <hobx_> try to hard, and overloaded network = more overloaded network ad infinum
[12:20] <toad_> eh?
[12:20] <toad_> if we do try too hard, overload causes both more overload and bad routing
[12:20] <dERMOTH> Caught java.lang.OutOfMemoryError running or seeding node
[12:20] <dERMOTH> java.lang.OutOfMemoryError
[12:20] <dERMOTH> Caught, in Main:
[12:20] <dERMOTH> java.lang.OutOfMemoryError
[12:20] <dERMOTH> java.lang.OutOfMemoryError
[12:20] <hobx_> Between plague and cholere, my choice is clear.
[12:20] <jay> greycat: does ftget calculate DBR's properly?
[12:20] <toad_> if we don't try too hard, then overload hopefully propagates back to the orig node, since we could probably do proper rate limiting
[12:20] <hobx_> toad: Yes, my point exactly.
[12:20] <toad_> i.e. propagate the rate limiting
[12:20] <dERMOTH> let'S try the old jvm...
[12:21] <toad_> make sure in queries >= out queries
[12:21] <toad_> dERMOTH: just increase the memory limit
[12:21] <greycat> jay: hmm, I don't know if I've ever tried ftget on a DBR site
[12:21] <toad_> -Xmx192M
[12:21] <hobx_> no
[12:21] <hobx_> screw that
[12:21] <toad_> hobx_: why?
[12:21] <jay> dERMOTH: try the old JVM with the same seednodes file
[12:21] <hobx_> it makes zero sense that in queries should equal out queries locally.
[12:21] <toad_> i'm not going back to initial-QR-95%-of-queries
[12:21] <jay> greycat: i wrote a DBR calculate function and i can't tell if it's correct
[12:21] <hobx_> And globally it happens as a matter of logic whatever you do.
[12:21] <toad_> hobx_: it enables us to propagate RNFs back to the source of the queries
[12:22] <toad_> thus if some idiot makes loads of requests, he gets RNFs - rather than the whole network getting RNFs
[12:22] <greycat> jay: are you doing anything with non-86400-second granularities?
[12:22] <dERMOTH> jay: trying.....
[12:22] <jay> greycat: in C there isn't a function to mktime() for UTC, only local
[12:22] <hobx_> your other idea to simply treat an RNF like a DNF and terminate the query was better.
[12:22] <toad_> hobx_: well, that's what I've been trying to tell KenMan, but he doesn't believe me :). Also there IS the issue of propagation.
[12:22] <jay> greycat: i don't think so
[12:22] <toad_> KenMan: why should we propagate quereis again?
[12:22] <greycat> jay: well that should keep it relatively simple, then
[12:22] <jay> greycat: i glommed a piece of code that did mkgmtime() and used C funcs
[12:23] <jay> greycat: it should
[12:23] <dERMOTH> toad_: while you were away, i rm -rf freenet then reinstaller both freenet and latest vm
[12:23] <KenMan> toad : hard numbers... first half of this hour, 265 queries hit exhaustion (RNF due to no routes), 680 came in .
[12:23] <jay> greycat: the date-code value is set at 0:00am for that day ?
[12:23] <hobx_> I think the problem of trying to keep an idiot of making loads of requests is intractable. It is the nature of this network that it is a common.
[12:23] <greycat> jay: I'm not sure why you're using mktime at all, though.
[12:24] <toad_> hobx_: I'm not prepared to give up on things as easily as you are. This can be demonstrated rather easily ;)
[12:24] <greycat> jay: time() returns the time in seconds-since-epoch which is naturally UTC, and then you just subtract off the time mod 86400....
[12:24] <toad_> KenMan: you were saying?
[12:24] <toad_> okay, 265 + 680 incoming
[12:24] <greycat> add an integer multiple of 86400 as needed to get the right day, then printf %x
[12:24] <toad_> no local traffic
[12:24] <toad_> how many outgoing?
[12:24] <toad_> 945 incoming that hour
[12:25] <jay> greycat: i was gmtime() to fill a struct tm and modified it that way
[12:25] <toad_> dERMOTH: hack the start-freenet.sh, increase the -Xmx
[12:25] <toad_> that's the memory limit
[12:25] <toad_> I would fix it with some horrible hack or other but I'm busy talking to KenMan about more strategic issues
[12:25] <jay> greycat: time % 86400 is the number of days then (minus the fractional day) ?
[12:25] <greycat> jay: when I test stuff by hand, I use code like this: perl -e '$t=time(); printf("%x\n", $t - $t%86400);'
[12:26] <KenMan> toad, no 265 out of 680
[12:26] <jay> greycat: oh ok
[12:26] <jay> i get it
[12:26] <dERMOTH> toad_: I'm first checking if the old jvm is working better... I know how to fix that
[12:26] <hobx_> When I say I think it is intractable, I mean that it seems to me logically that it cannot be done.
[12:26] <toad_> hobx_: okay, either we try to make incoming queries = outgoing queries by rate limiting... or we try to get NGRouting to realize that we suck, and we RNF 50% of our queries
[12:26] <jay> greycat: i meant to *say* that (perl code)
[12:26] <toad_> KenMan: 680 queries incoming
[12:26] <dERMOTH> same, java.lang.OutOfMemoryError
[12:26] <toad_> 265 of them RNFed IMMEDIATELY
[12:27] <KenMan> yes
[12:27] <toad_> that's nearly 40%
[12:27] <toad_> that's not nice
[12:27] <hobx_> I think you can't get a "keep stupid guy" control that works better then simply putting a limit in the source and hoping that few stupid guys can't change it.
[12:27] <toad_> have you added a proper diag for it?
[12:27] <KenMan> no, it really isn't nice. And I think you are getting hooked onto the problem now.
[12:27] * ninja_ (~im10ninja@3ffe:bc0:8000:0:0:0:0:3fab) Quit (Remote closed the connection)
[12:27] <iip_i2p> <Newsbyte> anyway, jay, don't feel to frustrated about the in your eyes trollish remarks of little me. Feel free to update the wiki any time! :-)
[12:27] <toad_> hobx_: we all know you're a quitter. that's not our problem.
[12:28] <KenMan> hobx - if stupid people circumvent the limits, they wind up damaging their anonymity :o
[12:28] <iip_i2p> <Newsbyte> btw, toad, I suggest you augment it to 15% and 320MB
[12:28] <toad_> Newsbyte: 15% of what?
[12:28] <toad_> KenMan: okay, step 1: make RNFs fatal
[12:29] <toad_> step 2: implement per node failure tables to prevent step 1 being exploitable
[12:29] <hobx_> toad_: As you wish. But think about it. Without global knowledge it cannot be done. The whole point with Freenet is that if you block somebody somewhere, the network routes around it.
[12:29] <iip_i2p> <Newsbyte> heh...of the free space of the HD, what else?
[12:29] <KenMan> step 3: watch everyone grow big grins on their faces
[12:29] <dERMOTH> thw weird part is that when I upgraded to 5090, i removed all routing files and had the seednodes from freenetproject... The only difference I can note is the 300 maximum threads instead of the 120? (I left the default) right now...
[12:29] <toad_> hobx_: well, you don't know much about rate limiting, right?
[12:29] <toad_> step 3: should we do something about local requests?
[12:29] <hobx_> Enforcing in=out locally makes zero sense to me. I can't see a single reason that would help anything.
[12:30] <toad_> hobx_: it would propagate load. We tell downstream nodes how many requests we can route.
[12:30] <toad_> then they send us no more than that
[12:30] <toad_> that will then propagate back to the request source
[12:30] <KenMan> toad: 3. probably . Yes, let's hit all the identified troubles at once, and call it 0.6 (after the grins subside anyway)
[12:30] <toad_> if he's spamming requests, HE and nearby nodes, get RNFs, rather than everyone getting RNFs
[12:30] <toad_> KenMan: sure :|
[12:31] <KenMan> toad: no. "sure :)"
[12:31] <toad_> KenMan: well, what do we need to do to prevent attacker easily seeing that most of our outgoing requests are locally originated?
[12:31] <toad_> I think the problem with Frost is insufficient NIO
[12:32] <toad_> it's not a problem if you're not running Frost, so I never got around to it
[12:32] <hobx_> There are no "nearby" nodes. When he consumes where he should be routing, he ends up routing elsewhere, but if the network works it comes back where it should go.
[12:32] <toad_> it might be a bug to do with timeouts
[12:32] <KenMan> rate control them. Queue them in FCP before searching, only let out 1/quota.
[12:32] <toad_> eh?
[12:32] <iip_i2p> <Newsbyte> so 15 instead of 10 ins't that dramatic that it could pose problems, yet it might help by making it bigger anyway
[12:32] <toad_> 1 in every quota?
[12:33] <toad_> Newsbyte: 15% of WHAT?
[12:33] <KenMan> as in, queue them up, dole them out at 200ms per query
[12:33] <KenMan> something along those lines
[12:33] <toad_> KenMan: then it'll just timeout and retry
[12:33] <dERMOTH> toad_: do you have any advice on using latest j2se 1.4.2 (05) on linux? I currently have latest 1.4.1
[12:33] <KenMan> that's problem 3 on our list of 3 priorities...
[12:34] <iip_i2p> <Newsbyte> I already said so!! The free space of your HD! Man, don't you remember your own posts on tha maillist?
[12:34] <toad_> well, somewhere around 2 would be: propagate request load
[12:34] <jay> dERMOTH: is it working?
[12:34] <jay> dERMOTH: the *latest* J2SE is 1.5.0 beta something
[12:34] * toad_ can't see rate limiting working _as originally designed_ without propagation
[12:34] <KenMan> yeah, that should help significantly. And no one but you and me (and mister black hat) see the severity of 3...
[12:35] <dERMOTH> right now I'm still with 1.4.1, usine |ux's seednodes.ref
[12:35] <dERMOTH> right now I'm still with 1.4.1, using |ux's seednodes.ref
[12:35] <KenMan> wait, i'm not sure I understand what you mean by propagation, but that's okay for now. I'm just glad you see the 40% hit exhaustion as a big problem.
[12:35] <toad_> KenMan: example of difficulties with 3.: local requests generally have a different success ratio than external requests
[12:35] <iip_i2p> <Newsbyte> keep it that way, der or move to 1.5 ;-)
[12:35] <toad_> yes, of course
[12:36] <toad_> KenMan: anyway there are more immediate issues
[12:36] <jay> dERMOTH: well im running debian linux on x86 using 1.4.2_05 no prob.. toad and kenman are obviously busy right noqw
[12:36] * toad_ thinks:
[12:36] <toad_> As a single unit:
[12:37] <toad_> - Make ALL QueryRejected's fatal (and hope looping isn't too much of a problem)
[12:37] <toad_> - Implement per node FTs
[12:37] <toad_> -- Release and reset unstable
[12:37] <toad_> right?
[12:37] <toad_> what DO we do about looping btw?
[12:37] <KenMan> party like it's 1999 also
[12:37] <toad_> KenMan: I judge by results
[12:37] <toad_> I'll believe it when I see it
[12:37] <KenMan> just measure it, for starters. I think the good will vastly outweigh the bad.
[12:38] <greycat> I'd rather party like it's 2005 and freenet works and there's no monkeys running my country
[12:38] * dERMOTH is re-trying 2.4.2
[12:38] * dERMOTH is re-trying 1.4.2
[12:38] <jay> dERMOTH: did 1.4.2_05 crash for you? if so what kind of crash?
[12:38] <dERMOTH> it was the outofmemerror
[12:39] <jay> dERMOTH: how much ram u have?
[12:39] <KenMan> per-route failure tables don't need to keep track of HTL, if that makes it simpler
[12:39] <dERMOTH> not physical ram, virtual ram
[12:39] <iip_i2p> <Newsbyte> hehe...ah well, your choice. if it gives probs (or not) feel free to update the wiki about it
[12:39] <jay> dERMOTH: try giving java more mem with the -Xmx switch
[12:39] <dERMOTH> the ram allocated by the jvm
[12:39] <toad_> KenMan: why do retries on RNFs cause more problems than additional hops?
[12:39] <jay> dERMOTH: yup i know
[12:40] <jay> dERMOTH: i start my node with 256M guaranteed to the VM
[12:40] <KenMan> i don't know the answer, and even if I did, I would never be able to communicate it to you :)
[12:40] <toad_> KenMan: you still have not explained why this happens, how die-on-RNF would help matters
[12:40] <KenMan> oh, wait. I do know this one.
[12:40] <jay> dERMOTH: in the end it uses less than half of what i guarantee java
[12:40] * dERMOTH doesn't have enough fingers to count the number of time he's been told tu use the -Xmx switch
[12:40] <KenMan> RNFs *cause* route exhaustion
[12:40] <toad_> KenMan: huh?
[12:40] <KenMan> rather, the retries do it
[12:40] <jay> dERMOTH: so what's it set at?
[12:40] <toad_> how?
[12:41] <KenMan> ins = outs philosophy ?
[12:41] <toad_> hobx_: okay, I admit that incoming requests = outgoing requests globally, as a mathematical fact
[12:41] <KenMan> believe me, it does. I don't have the time to make another graphic :(
[12:41] <toad_> hobx_: but overload causes us to attempt more outgoing requests than we should, globally
[12:41] <toad_> and they just RNF
[12:41] <jay> dERMOTH: the VM may be crashing because it doesn't have enough memory initially to parse the 24M seednodes file
[12:41] <iip_i2p> <Newsbyte> so, jay, as a coder and all that, I'm curious; in what area do you help with freenet, exactly? main javacode, GUI, encryption,..?
[12:41] * dERMOTH is so soncerned about what'S lest for his web browser that he prefer using a smaller seednodes ref than giving java more ram :)
[12:41] <toad_> KenMan: why do retries cause problems?
[12:41] <dERMOTH> what'S left
[12:42] <toad_> dERMOTH: everything
[12:42] <KenMan> they eat up the roadway needed for those 40% hitting the brick wall
[12:42] <toad_> KenMan: the request will finish eventually
[12:42] <jay> dERMOTH: well that's why it's crashing.. it is crashing on startup yes?
[12:42] <dERMOTH> not anymore
[12:42] <toad_> KenMan: only by compensating by dying on other nodes surely?
[12:43] <jay> Newsbyte: routing.. just look at the discussion between toad and kenman how difficult it is to make good routing decisions
[12:43] <toad_> I mean, why is it more of a problem (globally) for a node to go 4 hops and then instant-RNF
[12:43] <dERMOTH> i'm using |ux's seednodes.ref
[12:43] <jay> dERMOTH: ah ok
[12:43] <toad_> than for it to go 6 hops and succeed?
[12:43] <KenMan> don't you see the trouble with route exhaustion to begin with ? It is bad stuff, man !
[12:43] <toad_> or 6 hops and DNF?
[12:43] <toad_> I see the problem with route exhaustion
[12:43] <jay> dERMOTH: tell me the output to java -version
[12:43] <toad_> I'm just not sure I see how failing on RNF would improve the situation
[12:44] <dERMOTH> java version "1.4.2_05"
[12:44] <dERMOTH> Java(TM) 2 Runtime Environment, Standard Edition (build 1.4.2_05-b04)
[12:44] <dERMOTH> Java HotSpot(TM) Client VM (build 1.4.2_05-b04, mixed mode)
[12:44] <jay> Newsabte: routing is the brain behind Freenet.. everything else is built upon it
[12:44] <toad_> I suppose it would reduce the number of nodes touched by each request
[12:44] <toad_> but wouldn't reducing the max HTL do the same thing?
[12:44] <toad_> what's special about not retrying?
[12:44] <jay> dERMOTH: and what -Xmx is being passed to java ?
[12:45] <KenMan> okay. I have this request. I forward it, and get an RNF. So I retry. And get another RNF. And retry, and RNF. Oops, now I hit route exhaustion. Except I used up 3 slots worth of quota on a single request.
[12:45] <toad_> in particular, dying on what will remain very frequent, random events (such as looping), looks quite destructive
[12:45] <iip_i2p> <Newsbyte> yesyes, it's also the thing that sucks most for the longest time, no ;-) But, hey, I'm mildly optimistic :-)
[12:45] <toad_> KenMan: whereas if it spent the same HTL on 3 different nodes, it would still have used up 3 slots worth of quota
[12:45] <toad_> globally
[12:45] <KenMan> meantime, another request came in. It hit the wall immediately, thus helping some other node to get screwed up.
[12:46] <toad_> on node 1, route to node 2, used up 1 htl, on node 2, route to node 3, used up 2 htl...
[12:46] <jay> dERMOTH: i have no advice until you tell me what -Xmx is set to.. u know where to look?
[12:46] <jay> check start-freenet.sh
[12:46] <dERMOTH> java -Xmx128m -XX:MaxDirectMemorySize=128m freenet.node.Main
[12:46] <toad_> dERMOTH: increase the -Xmx
[12:46] <toad_> for the 27th time
[12:46] <dERMOTH> jay i'm a sysadmin
[12:47] <jay> dERMOTH: ok im a software developer and freenet hacker
[12:47] <jay> dERMOTH: freenet is memory intensive
[12:47] <toad_> KenMan: what were you saying?
[12:47] <toad_> 3 instant RNFs uses 3 HTL
[12:47] <KenMan> nothing, probably.
[12:47] <toad_> 3 hops uses 3 HTL
[12:47] <jay> dERMOTH: if you don't increase the size to at least *see* what is happening, we have no advice
[12:47] <toad_> why is the former more damaging than the latter?
[12:47] * dERMOTH said that because he'g getting lots of newbie-proof answers
[12:47] <jay> we know the Freenet code pretty well
[12:48] <toad_> dERMOTH: I have told you at least 5 times to increase the -Xmx
[12:48] <toad_> please try that!
[12:48] * dERMOTH also said it's working
[12:48] <toad_> ah
[12:48] <toad_> okay
[12:48] <jay> dERMOTH: well they just listen to everyone else
[12:48] <dERMOTH> lol
[12:48] <KenMan> FreeNet is attempting to perform Q multiplication, and as we have seen, globally that is impossible. Eventually the limits of the universe close in on a single node to suffocate it.
[12:48] <toad_> KenMan: no, it's bounded
[12:48] <toad_> KenMan: by the HTL
[12:48] <KenMan> nonsense
[12:48] <toad_> either we spend 3 HTL on routing through 3 hops
[12:48] <jay> dERMOTH: erm u said that it was crashing on memory
[12:49] <toad_> or we spend 3 HTL on 3 instant RNFs
[12:49] <toad_> either way we spend 3 HTL, and it's 3 htl closer to DNF
[12:49] <toad_> what's the difference? what's the problem?
[12:49] <KenMan> if you are telling me that you are okay expending all your outgoing query quota on 5% of all requests (and reject the other 95 immediately) then okay
[12:49] * dERMOTH said 2 or 3 times that he used |ux's seednodes.ref to fix it
[12:49] <toad_> KenMan: no, I'm trying to see why it makes ANY difference globally
[12:49] <jay> dERMOTH: that's not the issue.. we know that works
[12:50] <dERMOTH> well thanks for your help :)
[12:50] <KenMan> well, do you or don't you see why it makes a different locally ?
[12:50] <jay> dERMOTH: we'd like to fix the problem for memory errors on the large seednodes file
[12:50] <jay> dERMOTH: so there's no answer to what happens if you increase the memory passed to java then ??
[12:50] <Kyon> there an equivalent way to send -xmx to java, increasing memory in win32? i know very little of java, but freenet.node.main as a param does nothing, does the systray exe take params?
[12:50] <toad_> KenMan: well, it will make some difference, in that unsuccessful requests will go less far
[12:50] <toad_> however, it will also cause more unsuccessful requests
[12:50] <jay> Kyon: check flaunch.ini in the freenet dir
[12:51] <KenMan> There's only one way to test it out ... :)
[12:51] <jay> dERMOTH: *and* use the large seednodes file ?
[12:51] <toad_> KenMan: it's not that easy
[12:51] <KenMan> it never is
[12:51] <Kyon> jay: something like JavaMem=512m then?
[12:51] <toad_> KenMan: you have to convince ME before I inflict it on the rest of the network
[12:51] <KenMan> no, i don't.
[12:51] <toad_> yes, you do
[12:51] <jay> Kyon: heh i think so ;) i've never done it on windows myself
[12:52] <dERMOTH> as I said I preffer a smaller seednodes than giving too much memory to java... I only have 256mb and i'm running plenty of services, and sometimes I even run a 2nd desktop on another display
[12:52] <toad_> or I won't implement it, and I won't support your effort to reset the network
[12:52] <KenMan> not convincing you is always an option
[12:52] <KenMan> in fact, a frequent one
[12:52] <toad_> well fine, then freenet just won't work :)
[12:52] <KenMan> i only have myself to blame for poor communications
[12:52] <toad_> well
[12:52] <Kyon> jay: trying to get the large seed file to work as well, let me see
[12:52] <jay> dERMOTH: ok then that's cool.. i've helped a few people with this same issue
[12:52] <toad_> why is it better to make QRs fatal than to e.g. reduce the max HTL by 5?
[12:53] <toad_> why would the former work even though the latter doesn't? (it's been tried...)
[12:53] <jay> dERMOTH: im mainly concerned with fixing it so that the node can handle tons of seednodes
[12:53] <KenMan> the problem is local. The problem is route exhaustion. The problem is some queries causing others to hit exhaustion unfairly.
[12:53] <jay> Kyon: it works for me but i have 1G ram and -Xmx256 to Fred
[12:53] <toad_> jay: doing that consists of either selecting the seednodes by a php, or some incredibly hackish system to reduce the memory cost of selecting random seednodes
[12:54] <jay> toad_: i liked grecat's idea of the PHP seed generator
[12:54] <toad_> jay: a simple example: keep them as strings rather than as FieldSets and NodeReferences, until you need them
[12:54] <KenMan> the problem is an attempt to multiply queries in an environment where, ultimately, globally, religiously, that is impossible.
[12:54] <toad_> KenMan: no, the problem is global
[12:54] <toad_> the problem is that it happens on EVERY node
[12:54] <toad_> if it was just your node, I'd tell you where to shove it
[12:54] * dERMOTH whisper why he's "the one"
[12:54] <KenMan> well, i agree that people globally seem to report problems :)
[12:54] <dERMOTH> well ok
[12:55] <dERMOTH> let's try
[12:55] <jay> dERMOTH: ok so i dont have a real answer
[12:55] <jay> dERMOTH: other than what u can't do
[12:55] <toad_> KenMan: no. globally we do multiply queries. each query goes to 20 nodes.
[12:55] <dERMOTH> i meant wonder, not whisper
[12:55] <toad_> on each node, it blocks an MRI slot
[12:55] <KenMan> and each node takes in 20 times of queries
[12:55] <dERMOTH> just let me backup everything again :)
[12:55] <jay> dERMOTH: this issue is relatively new as far as i know
[12:56] <toad_> yes, incoming traffic for a node, on average, should be around 20 times total originated requests for a node
[12:56] <KenMan> well...
[12:56] <KenMan> that should perhaps be a limit...
[12:56] <toad_> KenMan: there is no possible way of enforcing it
[12:56] <leex> :) freenet is working fst today
[12:56] <toad_> except perhaps for propagating load
[12:56] <KenMan> i know i know
[12:56] <leex> norRNF
[12:56] <Kyon> jay: yeah that did it, changing javamem from default to 512m (i have 512m physical)
[12:56] <toad_> if we can find some way to do that
[12:56] <leex> hmm
[12:57] <jay> Kyon: i think 128M is even enough, 256M perhaps too much
[12:57] <KenMan> leex RNF is like a group of black goblins, that crawls around the network, and afflicts some cluster of nodes at a given time
[12:57] <Kyon> jay: yeah, just for test
[12:57] <jay> Kyon: if it temporarily reaches the limit will it crash the OS... ;)
[12:57] <KenMan> occasionally you will have good (RNF-free) results
[12:57] <jay> Kyon: cool
[12:57] <Kyon> jay: naw, windows has this cool 'paging' feature, ;-)
[12:58] <toad_> KenMan: afaics it might have a similar effect to reducing the max HTL slightly
[12:58] <jay> Kyon: feature ?
[12:58] <jay> heh
[12:58] <toad_> i.e. worse psuccess, but lower load
[12:58] <toad_> perhaps the lower load compensates for the worse psuccess
[12:58] <toad_> or perhaps it doesn't
[12:58] <Kyon> jay: just kidding.. sorry i am not a fan of 'dowz
[12:59] <toad_> on large networks, the worse psuccess is a problem; on small ones the lower load is important
[12:59] <KenMan> well, you have two suggestions (or was it 3) for things to try, based on some poorly-communicated logic
[12:59] * tyesinclair (~tyesincla@170.252.64.1) Quit (Read error: 104 (Connection reset by peer))
[12:59] <dERMOTH> jay: guess what? it's working now!! dammit
[12:59] <toad_> queueing might be an option, but it's going to be horrendously difficult to get it working without causing problems
[12:59] <jay> dERMOTH: heh.. what did u do?
[12:59] <toad_> based on our previous conversations re queueing
[13:00] <dERMOTH> I guess it didn't reseed... just let me try something...
[13:00] <KenMan> i feel that queueing is a weak reaction to inadequate rate limiting, but it may well provide a terrific benefit. Just seems like the harder path to me.
[13:00] <dERMOTH> [root@dermoth Freenet]# ./stop-freenet.sh
[13:00] <dERMOTH> [root@dermoth Freenet]# touch seednodes.ref
[13:00] <dERMOTH> [root@dermoth Freenet]# ./start-freenet.sh
[13:00] <leex> KenMan just its never been this fast before i getting 32k some times
[13:00] <jay> heh root?
[13:01] <dERMOTH> i don't care
[13:01] <leex> could be from local tho
[13:01] <toad_> KenMan: I feel that killing requests on the first loop or other QR is just going to make things worse
[13:01] <toad_> KenMan: is there a third option?
[13:01] <jay> dERMOTH: good thing the node doesn't do scripting
[13:01] <KenMan> fair enough. local balancing against remote traffic was #3 ?
[13:01] <dERMOTH> jay: now you know you can add a trojan that gain root access over here :)
[13:01] <toad_> well last time we tried, it caused a freezedown
[13:02] <toad_> it's just possible that last time we still had the RNF bug
[13:02] <jay> dERMOTH: easier said than done of course ;)
[13:02] <toad_> in which case we should definitely try again
[13:02] * toad_ checks...
[13:03] * toad_ wonders why he does what he does
[13:03] <leex> heh
[13:03] <dERMOTH> hey a hacker almost succeded inserting a backdoor in the linux kernel few month ago
[13:03] * KenMan 's primary discontent with rate limiting is that route exhaustion can occur, and so frequently
[13:03] <leex> coes toad_ isgood at it and loves doing it
[13:03] <toad_> KenMan: okay so how do we prevent it?
[13:03] <toad_> leex: am I?
[13:03] <toad_> I'm skeptical that anyone is
[13:03] <dERMOTH> he got caught by the paranoid check on the cvs
[13:04] <leex> heh
[13:04] <toad_> okay, we last tried it on build 6476
[13:04] <KenMan> my brain's routes are currently 100% exhausted :(
[13:05] <toad_> that was way before the recently fixed RNFBug was introduced
[13:05] <hirvox> dermoth:but that cvs was read-only. no developer uses it to commit stuff. Normal kernel development is done with bitkeeper
[13:05] <toad_> however, there may have been some multiplication of requests going on at the time..
[13:06] * Hory (~Miranda@81.196.25.110) Quit (No route to host)
[13:06] <dERMOTH> yes but that backdoor could have commited by a trusted developper using CVS
[13:06] <KenMan> i feel that RNFs are both causing multiplication and resulting from it
[13:06] <dERMOTH> yes but that backdoor could have been commited by a trusted developper using CVS
[13:07] <toad_> KenMan: quite probable
[13:07] <KenMan> is that the definition of 'feedback' ?
[13:07] <toad_> yes
[13:07] <KenMan> ;)
[13:08] * toad_ thinks that when we tried last, we had problem precisely because we still had explicit request multiplication
[13:08] <hirvox> dermoth:no, that CVS couldn't have been used to commit anything. it was a read-only copy of the main bitkeeper tree, so that backdoor was nowhere near the main kernel tree
[13:08] <toad_> i.e. we didn't decrement HTL on retrying after a QueryRejected (!!)
[13:08] <toad_> if(accepted)
[13:08] <toad_> hopsToLive = Math.min(hopsToLive-1, qr.getHopsToLive());
[13:08] <toad_> aha
[13:09] <toad_> okay the last time we tried it, we DID reduce HTL when we got a proper RNf
[13:09] <toad_> but NOT when we got an early QR
[13:09] <toad_> that might have been enough
[13:09] <toad_> to break it
[13:09] <toad_> also...
[13:09] <toad_> we were trying to optimize routing
[13:09] <toad_> not just to eliminate RNFs
[13:09] <toad_> okay
[13:10] <toad_> I can code up in <= out relatively easily
[13:10] <toad_> testing it will be more interesting... perhaps I can do it on a testnet?
[13:11] * Hory (~Miranda@81.196.25.110) has joined #FreeNet
[13:11] <dERMOTH> toad_: doesn'T touch'ing seednodes.ref force reseeding?
[13:11] <toad_> dERMOTH: should o
[13:11] <toad_> do
[13:11] <toad_> if the node has been shut down
[13:11] <dERMOTH> i did
[13:12] <dERMOTH> jay: forcing a reseeding doesn'T seem to trigger a oom...
[13:12] <dERMOTH> i'll try removing rooting tables
[13:12] <dERMOTH> routing tables
[13:13] <toad_> KenMan: actually, it looks like the last time we tried we were trying just to make incoming == outgoing...
[13:13] <dERMOTH> jay?
[13:14] <jay> dERMOTH: yeah lunch
[13:18] <toad_> hmmm
[13:18] <toad_> maybe I see what you're getting at...
[13:18] <KenMan> hell, i hope so. I need to go back to sleep for an hour or two now...
[13:18] <toad_> well
[13:18] <toad_> 40% of requests die immediately
[13:18] <dERMOTH> still working
[13:18] <toad_> lets say the remaining 60% cause 120% traffic
[13:18] <toad_> now the obvious propagation logic would be:
[13:19] <dERMOTH> let's remove node too
[13:19] <toad_> load = outgoingRequests / incomingRequests
[13:19] <toad_> the problem with that would be: does rate limiting influence the ratio?
[13:19] <toad_> I suppose maybe it does?
[13:19] <KenMan> rate limiting is both too fine-grained and yet not accurate enough at the same time.
[13:20] <toad_> KenMan: okay
[13:20] <toad_> well
[13:20] <toad_> if we have a 300% ratio for the above
[13:20] <toad_> and we reduce our traffic by 70%
[13:20] <toad_> will that result in a 100% ratio?
[13:20] <toad_> on average?
[13:20] <KenMan> and to answer an earlier quesiton you had - isn't working towards a limit the same thing as an average? No! it is not.
[13:21] <KenMan> you just can't allow such a high percentage of queries to replicate
[13:22] <toad_> okay, I have made appropriate changes
[13:22] <KenMan> when you work towards a limit, you hit a brick wall. when you work towards an average, the wall experiences some weathering, and simply represents an average
[13:22] <toad_> lets see what happens...
[13:23] <toad_> first on the regular network
[13:23] * toad_ removes the rate limiting data files so that it reacts faster...
[13:23] <dERMOTH> jay: still working...
[13:23] * hobx (~hobx@h196n1fls21o1077.bredband.comhem.se) has joined #freenet
[13:23] * ChanServ sets mode +o hobx
[13:24] * toad_ grrr, takes ages to load filesyste
[13:24] <toad_> m
[13:25] * KenMan thinks that route exhaustion probably should be made into a proper diagnostic...
[13:25] <KenMan> but later.
[13:25] <KenMan> sleep now.
[13:25] * toad_ thinks so too
[13:26] <KenMan> bb90
[13:27] <iip_i2p> <mule> what kind of traffic is included in bw limit calculations? the rate fproxy tells me is close to the bw limit set. the traffic on the network interface however is significantly higher (up to 3 times) with nothing else active.
[13:28] <iip_i2p> <mule> it seems the outgoing traffic rises with increased incoming traffic.
[13:28] <dERMOTH> toad there's a little bug on the start or preconfig script... looks like storeSize doesn'T take letters:
[13:28] <dERMOTH> storeSize [268435456]
[13:28] <dERMOTH> 3000M
[13:28] <dERMOTH> storeSize [268435456]
[13:28] <dERMOTH> 3000M
[13:28] <dERMOTH> storeSize [268435456]
[13:28] <toad_> dERMOTH: hmmm
[13:28] <dERMOTH> i've allways has storeSize=3000M in my freenet.conf
[13:30] <iip_i2p> <mule> the outgoing traffic on the network interface, the value reported by fproxy is unchanged.
[13:31] <toad_> hmmm
[13:31] <toad_> according to Open Connections:
[13:31] <toad_> Number of requests (sent/received) 58/109
[13:31] <toad_> yet according to general info the sent is much higher
[13:31] <dERMOTH> jay: caught OOM. I'll try with the previous freenet.cong before trying -Xms...
[13:34] <iip_i2p> <mule> toad_ was that meant for me? my reports on outgoing traffic in fproxy are taken from general info.
[13:36] <toad_> no
[13:36] <toad_> mule: there's a hard limit at 140%
[13:36] <toad_> i wouldn't expect it to go above that much
[13:36] <dERMOTH> Starting Freenet now: Command line: java -Xmx256m -XX:MaxDirectMemorySize=128m freenet.node.Main
[13:36] <dERMOTH> Done
[13:37] * dERMOTH whish himself good luck
[13:37] * dERMOTH freaks out as he see his mem usage going up
[13:39] <iip_i2p> <mule> toad_, i have outgoing traffic as reported by knetload in the range of up to ~200kBit/s, where my limit is 8kByte/s and the reports by fproxy are in that range.
[13:39] <toad_> 200kbps = 25kB/sec...
[13:39] <toad_> hmmm
[13:39] <toad_> mule: output or input limit?
[13:40] <iip_i2p> <mule> is the response to incoming traffic also considered for bw limiting?
[13:40] <dERMOTH> jay: It's working but fred still use a sh*tload of memory...
[13:40] <iip_i2p> <mule> output limiting.
[13:40] <jay> dERMOTH: indeed
[13:40] <jay> dERMOTH: looks like it's not the JVM per se but memory
[13:40] <toad_> mule: strange
[13:40] <toad_> last time I checked, output limiting seemed to work reasonably well
[13:40] <toad_> are you sure it's not something else?
[13:41] <dERMOTH> jay: looks like java didn't freed the memory:
[13:41] <dERMOTH> Maximum memory the JVM will allocate 260,160 KiB
[13:41] <dERMOTH> Memory currently allocated by the JVM 231,552 KiB
[13:41] <dERMOTH> Memory in use 163,202,664 Bytes
[13:41] <dERMOTH> Estimated memory used by logger None
[13:41] <dERMOTH> Unused allocated memory 73,904,608 Bytes
[13:42] <toad_> hmmm... requests are getting multiplied even though there is little backoff
[13:42] <iip_i2p> <mule> my input limit is set to 64000Byte/s. and yes, if i stop freenet the load is gone.
[13:42] <toad_> probably caused by RNFs on other nodes...
[13:42] <dERMOTH> and it's painfully slow :(
[13:42] <toad_> it MAY be falling...
[13:42] <toad_> in which case it would just be due to startup... hmmm
[13:43] <jay> dERMOTH: mine is only holding 81M
[13:43] <dERMOTH> at least unused memory finally got swapped out and it's going better
[13:43] <dERMOTH> 25298 root 15 10 191M 108M 6984 S N 13.5 43.1 0:03 java
[13:43] <jay> but ive been running my node for more or less a week now
[13:43] <dERMOTH> yeah I know it's root
[13:43] <dERMOTH> lol
[13:44] <jay> dERMOTH: the environment page (in advanced mode) give better info on mem usage
[13:44] <jay> from the gateway
[13:44] <toad_> mule: also, it should eventually level out because of rate limiting
[13:44] <toad_> mule: do you have any wierd config settings?
[13:44] <dERMOTH> jay: yes it's up here ^^^^^^
[13:44] <toad_> there are config settings to disable low level limiting...
[13:45] <jay> dERMOTH: ya saw it
[13:45] <iip_i2p> <mule> not that i know. and the report in fproxy nicely follows my settings.
[13:45] <toad_> hmmm
[13:45] <toad_> 22/39 backed off...
[13:46] <toad_> mule: I thought you said the report in fproxy was accurate?
[13:46] <dERMOTH> so you still need me or I can get back my 20-nodes-ref-30-meg-of-ram-freenet-node?
[13:46] <iip_i2p> <mule> sorry, my wife calls for dinner. back soon.
[13:46] <jay> dERMOTH: sure go ahead.. thanks for the effort
[13:46] <toad_> ok
[13:46] <jay> dERMOTH: i didn't think freenet ate up that much memory
[13:47] <jay> from the seeds that is
[13:47] <dERMOTH> Memory in use 161,775,904 Bytes
[13:48] * toad_ starts testbed using the new code...
[13:49] <jay> dERMOTH: it should fall much more than that i would think.. but it's falling nonetheless
[13:51] <dERMOTH> well it's been more than a week that I resurected my node and over time it raised rather than lowered.
[13:51] <dERMOTH> see now:
[13:52] <dERMOTH> 25580 root 17 10 36260 35M 8572 S N 1.5 14.1 0:00 java
[13:52] <leex> i cant use the .ref file from freenets home page i get out of mem errors (in the freenet icon logs)
[13:52] <dERMOTH> Maximum memory the JVM will allocate 130,112 KiB
[13:52] <dERMOTH> Memory currently allocated by the JVM 17,876 KiB
[13:52] <dERMOTH> Memory in use 13,637,736 Bytes
[13:52] <dERMOTH> Estimated memory used by logger None
[13:52] <dERMOTH> Unused allocated memory 4,665,392 Bytes
[13:53] <jay> dERMOTH: using the smaller seednodes?
[13:53] <dERMOTH> yes
[13:54] <dERMOTH> and given the fact that I have enought free memory for my web browser it'S working better afert few minutes than my one week old node
[13:56] * kers (~kers@192.ppp141.rsd.worldonline.se) has joined #freenet
[13:56] <dERMOTH> so which settings I need to halve to keep half of my ref?
[13:57] <dERMOTH> rtMaxNodes and maxRoutingSteps?
[13:57] <dERMOTH> and maxNodeConnections?
[13:59] <toad_> ignore rtMaxNodes
[13:59] <toad_> what you want to do is simply delete half the ref file
[14:00] <dERMOTH> right now it contains only 20 nodes
[14:00] <dERMOTH> but I guess it will grow
[14:00] <toad_> okay so what's the proble,m?
[14:01] <jay> OOM exception on large seednodes file
[14:01] <jay> or with lots of mem, it runs badly
[14:01] <jay> since he has i think 256 physical ram (?)
[14:01] <dERMOTH> yes
[14:01] <dERMOTH> I want to reduce mem usage
[14:02] <jay> toad_: it runs badly with -Xmx256 for example
[14:02] <toad_> hmmm
[14:02] <dERMOTH> otherwise i can'T keep my node running over time, it clow down too much my system
[14:02] <toad_> dERMOTH: so set -Xmx192?
[14:02] <dERMOTH> it run badly even with 128
[14:03] <dERMOTH> because java itself use memory, so it higher
[14:03] <dERMOTH> what I want is permanently reduce the number of refs so it doesn't go too high on memory
[14:05] <dERMOTH> right now i have 20 refs and it's onty using ~30mb. Before that I had 192 refs and it was a nightmare, around 140m
[14:06] <jay> im not sure we have anything to limit *that*
[14:06] <jay> dERMOTH: but i see your point clearly
[14:07] <dERMOTH> see, it's now at 33 refs, and using more memory...
[14:08] * ninja_ (im10ninjas@82-133-66-19.dyn.gotadsl.co.uk) has joined #freenet
[14:08] <dERMOTH> it started at 20, in 15 minutes it weng to 33, and will grow too much if I don't put limits
[14:09] <dERMOTH> but I think it's not only a single setting. I believe you have to keep a good balance betweeb settings...
[14:14] * dERMOTH shout "HELLO", but only hear his echo back
[14:20] <iip_i2p> <mule> toad_, back again. if i look at the difference between transmitted bytes over 10 seconds with ifconfig, it reports 20kByte/s.
[14:21] <iip_i2p> <mule> and that seems to be my throughput limit of the uplink.
[14:21] <toad_> mule: yet on the General infolet, it says it's using 6kB/sec?
[14:21] <iip_i2p> <mule> Current upstream bandwidth usage 8306 bytes/second (103.8%)
[14:22] <iip_i2p> <mule> it's set to 8000Bytes/s
[14:22] <leex> woho me internet is running at 1500
[14:22] <leex> or 164 (rip offs)
[14:22] <leex> it should be 184
[14:23] <leex> 187.5KB
[14:23] <leex> :)
[14:23] <iip_i2p> <mule> may it be caused by my significantly different upstream/downstream?
[14:24] <leex> whats you internet speed
[14:24] <toad_> mule: no, when we tested it it was on an asymmetric link iirc
[14:25] <toad_> mule: how asymmetric?
[14:25] <toad_> grossly asymmetric like 10:1 ?
[14:25] <toad_> it's possible that would cause it...
[14:25] <iip_i2p> <mule> i don't exceed bw limits on another machine set to 50kByte/s, at least not significantly. it has 50k up, 50k down and is volume limited.
[14:26] <iip_i2p> <mule> like 8:1
[14:26] <toad_> tcp acknowledgement traffic is hard ot measure
[14:26] <leex> 1500/256
[14:26] <toad_> hmm
[14:26] <toad_> that might be it
[14:26] <toad_> try reducing your input limit
[14:26] <leex> i norm get 2-3 k over head with me line
[14:27] <iip_i2p> <mule> but right now input is low and output still high.
[14:27] <jay> toad_: can u recommend where i can insert a log command to print out the DBR target string on ClientGet (KSK@test-abc123) ?
[14:27] <jay> toad_: for either fproxy or the cli util
[14:28] <jay> DateRedirect.java ?
[14:28] <toad_> jay: hmm
[14:28] <toad_> try there
[14:28] <toad_> or just turn on enough client logging to see it follow the redirect
[14:28] <jay> toad_: what do i pass to logLevelDetail exactly?
[14:29] <toad_> okay, network load isn't very reliable... use acceptedExternalRequestCounter...
[14:29] <toad_> jay: logLevelDetail=freenet.client:debug would do it, but that's probably TMI
[14:29] <jay> ah i see
[14:29] <jay> works
[14:32] * toad_ loads up...
[14:33] <toad_> first time, it didn't work...
[14:33] <toad_> because of a stats bug
[14:33] <toad_> this time, it should work
[14:34] <toad_> this is bizarre
[14:34] <toad_> by fetching http://127.0.0.1:16309/
[14:34] <toad_> i can triggrer:
[14:35] <toad_> java.lang.NoClassDefFoundError: freenet/support/LRUQueue$ItemEnumeration
[14:35] <toad_> at freenet.support.LRUQueue.elements(LRUQueue.java:60)
[14:35] <toad_> at freenet.OpenConnectionManager.countOpenLRUConnections(OpenConnectionManager.java:518)
[14:35] <toad_> at freenet.node.Main.dumpInterestingObjects(Main.java:3757)
[14:35] <toad_> at freenet.thread.YThreadFactory$YThread.run(YThreadFactory.java:292)
[14:35] <toad_> even though it's completely unrelated
[14:38] <toad_> hmmm
[14:43] <toad_> was a problem with the node already being up
[14:44] <jay> / note that this must go AFTER DateRedirect, due to inheritence reasons. Otherwise, a big scary monster WILL EAT YOU.
[14:44] <jay> nice comment
[14:44] <greycat> heh
[14:45] <toad_> hmm
[14:45] <toad_> /
[14:46] <toad_> ahhh :)
[14:46] <toad_> root@amphibian:/usr/src/cvs# cat /root/seednodes.ref | grep version= | wc -l
[14:46] <toad_> 634
[14:46] <toad_> woah
[14:47] <toad_> 9 version=Fred,0.5,STABLE-1.50,5083
[14:47] <toad_> 625 version=Fred,0.5,STABLE-1.50,5084
[14:47] <toad_> heh
[14:47] <toad_> not fucking useful
[14:47] <toad_> shit
[14:47] <toad_> all the seednode source node ops didn't upgrade!
[14:47] <toad_> hmmm
[14:47] <lolo-laptop> Uptime 3 days 11 minutes
[14:47] * toad_ doublechecks
[14:47] <toad_> maybe that's an old one
[14:47] <lolo-laptop> Smoothed local mean traffic (queries per hour): 2892.550
[14:48] <toad_> okay
[14:48] <toad_> it has 666 refs
[14:48] <lolo-laptop> unique contacts: 2742, active connections: 0
[14:48] <toad_> of which 448 are 5090!
[14:48] <toad_> woah
[14:49] <toad_> lolo-laptop: woah, over what period?
[14:49] <lolo-laptop> Connections open (Inbound/Outbound/Limit) 165 (122/43/200)
[14:49] <lolo-laptop> toad_: read up
[14:49] <toad_> cool
[14:49] <toad_> even if they're all on 1-day-IP DSL's, that's 900 nodes...
[14:50] <toad_> oooh
[14:50] <toad_> this propagation code appears to be working on the localtestnet
[14:50] <iip_i2p> <mule> i fired up ethereal and see mostly small data packets in the range of 200bytes. length of ACKs being mostly 66 bytes, this means that if the tcp ack traffic for incoming is not considered in the outgoing bw, it is likely to eat a large part of the available upload bw. i further see frequent retransmits.
[14:50] <toad_> i'm going to need some volunteers soon for wider testing...
[14:50] <toad_> mule: 66 bytes? woah
[14:51] <toad_> i thought they were less than that
[14:51] <toad_> is that including the frame header?
[14:51] <toad_> okay
[14:51] <toad_> node loads, due to propagation:
[14:51] <toad_> 100%
[14:51] <toad_> 103.8%
[14:51] <iip_i2p> <mule> yes, all figures for "bytes on wire"
[14:51] <toad_> 38.7%
[14:51] <toad_> 111.2%
[14:51] <toad_> 129.6%
[14:52] <toad_> 100.1%
[14:52] <toad_> 18%
[14:52] <toad_> 88.6%
[14:52] <toad_> hrrm
[14:53] <toad_> 4 above 100%, 3 below... average is 86% !
[14:53] <toad_> cool!
[14:53] <toad_> mule: hrrm
[14:53] <toad_> well
[14:53] <toad_> what's the ethernet frame header? 16 bytes?
[14:54] <toad_> mule: unfortunately there's no way to know when either retransmits or acks occur
[14:54] <toad_> acks can be for several packets
[14:54] <toad_> generally, seriously asymmetric connections SUCK :(
[14:54] <toad_> HEY PEOPLE!
[14:54] <toad_> ANYONE WANT A NEW JAR!
[14:55] * toad_ needs some volunteers to set up a larger testnet, which may become new unstable
[14:55] <toad_> any takers?
[14:55] <verl> yes.
[14:55] * toad_ is being very, very careful after what happened last time...
[14:55] <lolo-laptop> toad_: I'm still running my node for uptime, so I'm going to sit this one out
[14:55] <toad_> lolo-laptop: uptime?
[14:55] <toad_> what do you mean
[14:55] <toad_> ?
[14:56] <toad_> lolo-laptop: "I've got the highest uptime of any freenet node anywhere!!!" or is it "I'm trying to figure out whether JVM XYZ crashes..."?
[14:56] <iip_i2p> <mule> ethernet 12 bytes, IP 20 bytes, tcp 32. figures vary for different ack frames, though.
[14:57] <lolo-laptop> toad_: the latter
[14:57] <toad_> mule: hmmm
[14:57] <toad_> I thought it was 24 for I~P
[14:57] <toad_> IP
[14:57] <toad_> at the moment we use 24...
[14:57] <toad_> what is it for a TCP packet over ethernet? 12+20+32? or is 32 for an ack?
[14:58] <iip_i2p> <mule> that were the fields listed in an ack in the trace
[14:58] <KenMan> 40+ for the header, no ?
[14:59] <toad_> mule: okay
[14:59] <toad_> how much for a data packet?
[14:59] <greycat> For whatever it's worth, I saw a great amount of ACK queue activity on my firewall when I had my node running with all of those active transfers. But I don't have numbers for you.
[14:59] <toad_> KenMan: want in on the propagation-of-load testnet?
[14:59] <KenMan> 20bytes IP, 20bytes TCP, assume no options
[14:59] <KenMan> sure, why not ?
[14:59] <toad_> KenMan: is that accurate or a rough guess?
[15:00] <KenMan> that's accurate for IP & TCP
[15:00] <toad_> KenMan: okay
[15:00] <leex> 40 is the max size for the header i think with options
[15:00] <toad_> what about acks?
[15:00] <leex> http://www.speedguide.net:8117/
[15:00] <leex> with or with out options
[15:00] <KenMan> you don't get to see ACKs from an application level, they occupy one bit.
[15:01] * jay (jcl@ool-18bf6dac.dyn.optonline.net) Quit (Remote closed the connection)
[15:01] <toad_> Your RcvWindow limits you to: 235.52 kbps (29.44 KBytes/s) @ 200ms
[15:01] <toad_> Your RcvWindow limits you to: 94.208 kbps (11.776 KBytes/s) @ 500ms
[15:01] <toad_> Consider increasing your RWIN value to optimize TCP/IP for broadband.
[15:01] * jay (jcl@ool-18bf6dac.dyn.optonline.net) has joined #freenet
[15:02] <toad_> i wonder if that's from the transparent proxy...
[15:02] <toad_> KenMan: yes but for bandwidth limiting purposes?
[15:02] <toad_> okay
[15:02] <toad_> node is u[p
[15:02] <toad_> up
[15:02] <toad_> who wants the jar?
[15:02] <KenMan> gimme gimme. I have no expectations.
[15:02] <leex> i should run unstable
[15:03] <toad_> hmm, just need to fix it not detecting its address :)
[15:03] <KenMan> oh yeah. What is it, unstable ? I'm seeded and all that for stable :(
[15:03] <iip_i2p> <mule> i have many frames with 226 bytes, of which 160 are data
[15:03] <toad_> KenMan: unstable reset
[15:04] <KenMan> oh, then it won't much matter.
[15:04] <leex> in my stable i sees mytranperan proxy server as 1 and 95 of my ip in stable
[15:04] <toad_> mule: 46 bytes overhead? woah
[15:04] <toad_> 66 bytes even
[15:04] <toad_> that's a good argument for keeping the padding size relatively high
[15:04] <toad_> KenMan: can you accept dcc or should i email you?
[15:05] <toad_> jay: want in?
[15:05] <KenMan> email me
[15:05] * verl wants
[15:05] <verl> (dcc)
[15:06] <iip_i2p> <mule> many packets are in the range 210-250 bytes, with 160 to 200 bytes of data, rest is overhead.
[15:06] <toad_> anyone else? won't be much of a testnet with only 3 nodes :)
[15:07] <leex> how i test
[15:07] <Kyon> i will
[15:07] <iip_i2p> <mule> btw, the overhead above was 66, not 46
[15:07] <leex> del every thing but store folder
[15:07] <toad_> leex: you run the jar and the seednodes.ref I send you
[15:07] <toad_> and do what I say, and generally test it
[15:07] <toad_> this is a small distributed testbed network
[15:07] <leex> k
[15:07] <toad_> so that we hopefully pick up if it's going to mess up the network if we release it
[15:07] <toad_> leex: can I dcc?
[15:07] * leex dcc
[15:07] <toad_> Kyon: can I dcc?
[15:07] <leex> heh
[15:08] <jay> toad_: i got the node to spit out DBR strings with the logger (working on dbr's in fcptools atm)
[15:08] <Kyon> toad_: not sure, give it a try
[15:08] <toad_> if it works, we'll use it as the core of a new unstable branch network
[15:09] <toad_> jay: cool!
[15:09] <verl> toad: we should delete routing table yes?
[15:09] <toad_> verl: yes
[15:09] <toad_> delete rt* lsnodes* ngrt_*
[15:09] * dERMOTH (~dERMOTH@dsl-131-244.aei.ca) Quit (Client Quit)
[15:09] <iip_i2p> <mule> so if you calculate 8k limit, this will be 10k on the wire. add 1/4 of 16k data input for acks, we are at 14k average output on the wire, with 8k output limit. that explains my overloaded uplink.
[15:09] <toad_> might as well keep the store intact
[15:10] <toad_> mule: well, we can increase the overhead assumption to 66 bytes
[15:10] <toad_> are you sure that's not just some wierdness of your link?
[15:10] <toad_> it seems like a lot :|
[15:10] <leex> how do i see what the packets are doing
[15:10] <toad_> leex: hmm?
[15:10] <toad_> oh, ask mule
[15:10] <leex> you lot are talking about over head
[15:11] <toad_> leex: use the jar, and myref.ref for seednodes.ref
[15:11] <toad_> KenMan: got the mail yet?
[15:11] <KenMan> no
[15:11] <iip_i2p> <mule> calculation wrong, the 16k input will give even more ack output.
[15:11] <iip_i2p> <mule> leex, are you on linux?
[15:12] <toad_> KenMan: 20 TCP + 20 IP + how much link-local? 26? 16?
[15:13] <KenMan> you mean, an ethernet header ??
[15:13] <toad_> mule gets 66 bytes total overhead experimentally...
[15:13] <toad_> yeah
[15:13] <toad_> usually
[15:13] * toad_ thinks PPP etc are less
[15:13] <toad_> wifi is probably more
[15:13] <KenMan> hmmm, not sure.
[15:14] <toad_> well lets go with 66 bytes for now
[15:14] <toad_> however absurd it may seem :|
[15:14] <iip_i2p> <mule> i'm on adsl with pppoe, but that's the traffic to the router.
[15:14] * toad_ wonders if we should queue packets for 500ms or so before sending them just to make sure we get a nice aggregation...
[15:14] <iip_i2p> <mule> so pppoe shouldn't be of importance
[15:14] <leex> no i am on windows
[15:14] <leex> mule
[15:15] <KenMan> looks like around 14-16 bytes for ethernet
[15:15] <toad_> verl, leex : got the nodes up yet?
[15:15] <verl> getting the 1st conn went faster this testnet :)
[15:15] <toad_> KenMan: so how come mule sees 66 total?
[15:15] <KenMan> tcp options
[15:15] <toad_> yay, got one
[15:15] * toad_ would like to have more... :)
[15:16] <iip_i2p> <mule> leex, i use a package called ethereal, don't know whether it is available for windows.
[15:16] <toad_> KenMan: they don't just happen when explicitly enabled at app level?
[15:16] <toad_> okay, initial spike occurring...
[15:16] <toad_> probably just due to initial CPU usage...
[15:16] <iip_i2p> <mule> as stated above: overhead for ethernet, ip, tcp add up
[15:16] <KenMan> no, they can happen 'on their own'
[15:17] <toad_> hmmm
[15:17] <toad_> lost a connection
[15:17] <toad_> oh i see
[15:17] <KenMan> where is your email to me ?? should be mostly across the atlantic by now :o
[15:17] <leex> should i change my ports for testing
[15:18] <KenMan> probably, yes leex
[15:18] <toad_> leex: yes, i haven't, and i have all these unconnectable nodes :)
[15:18] <toad_> KenMan: you are Ken Corson <jkcorson@comcast.net>, right?
[15:18] <KenMan> yeah, that's me
[15:19] <toad_> oh and i apologize for the spammers that just got hold of your email :)
[15:19] <KenMan> it just got here...
[15:20] <toad_> Load due to propagation = 2,508.2% = 690.3311464756864 / 27.522935779816514
[15:20] * leex sees if he is as smart as he thinks (fireing up freenet
[15:20] <toad_> hmmm
[15:21] <toad_> here's an interesting behaviour...
[15:21] * KenMan bids netstable farewell, it was brief but enjoyable...
[15:21] <toad_> when it first starts, the incoming reqs is too low
[15:22] <KenMan> well, sure. you wont have very many routes now, will you ?
[15:22] <toad_> Load due to propagation = 168.1% = 97.9005765263592 / 58.251484603809
[15:22] <toad_> hmmm
[15:22] <toad_> okay, make some requests...
[15:22] * toad_ makes some reqs
[15:22] <toad_> actually i'll insert something
[15:23] <toad_> CHK@rQbfYmj0zVnjgNKt7cQwKGWiTf0RAwI,Vrj9YFvW65xMDtdKZ8gz4A/moo.zip
[15:24] <Kyon> im up
[15:24] <leex> looks like its working
[15:24] <leex> 1 node cinnected
[15:24] <leex> connected
[15:24] * verl will download as soon as you are not backed off...
[15:24] <iip_i2p> <mule> is there any real control of incoming traffic? or is it only the based on reqs/h?
[15:25] <toad_> hmmm
[15:25] <toad_> 2 nodes have 10 minute MRIs...
[15:25] <toad_> mule: what do you mean?
[15:25] <Kyon> yeah backed off here as well
[15:25] <toad_> ahhh
[15:25] <toad_> falling...
[15:25] <toad_> what's my MRI?
[15:25] <toad_> i'm 82.something
[15:25] <KenMan> here goes nothing...
[15:26] <toad_> Load due to propagation = 206.1% = 65.08533410471507 / 31.586151227472932
[15:26] <toad_> well, something isn't working...
[15:26] <KenMan> i have one toadnode
[15:26] <Kyon> Load due to propagation = 35,256.4%
[15:27] <verl> Current estimated load for rate limiting 88,8%
[15:27] <iip_i2p> <mule> how can you control the number of bytes sent to you? only low level mechanism i'd see is small window size and not sending acks. how do you steer incoming traffic so it does not exceed input bw limits?
[15:27] <verl> Load due to propagation = 4,2%
[15:27] <KenMan> MRI for toadnode 642568
[15:27] <toad_> globalQuota: 160.84669419495347, totalRequests: 940.7175021535694, totalRequ
[15:27] <toad_> estsThisNode: 100.0, thisNodeMinQuota: -154.7560284409156, thisNodeQuota: -154.7
[15:27] <toad_> 560284409156, minRequestInterval: 600000.0
[15:27] <toad_> hmmm
[15:28] * KenMan will be patient then
[15:28] <iip_i2p> <mule> because with the large part of overhead, to limit outgoing traffic needs to precisely limit incoming traffic.
[15:29] <toad_> Load due to propagation = 667.4% = 210.7922727339275 / 31.586151227472932
[15:29] <toad_> hmmm
[15:29] <leex> i keep on getting an backed of node
[15:30] <toad_> hmmm
[15:30] <toad_> how can it be backed off until 1361ms, when the MRI is only 600?
[15:30] <toad_> averages...
[15:30] <toad_> hmmm
[15:31] * Diablo-D3 (~diablo@dialup-4.156.102.183.Dial1.Boston1.Level3.net) has joined #freenet
[15:31] <Diablo-D3> hey guys
[15:31] <KenMan> hey, what do you think of D3 ?
[15:32] <leex> looks nice
[15:32] <Diablo-D3> I dont know, seeing as my nick is not game related
[15:32] <leex> i played it an little
[15:32] <leex> lol
[15:32] <KenMan> who said I thought it had anything to do with your nick ? sheesh ! maybe I just wanted to contribute to the OT traffic in here !
[15:33] <Diablo-D3> KenMan: ...
[15:33] <Diablo-D3> Anyhow
[15:33] <Diablo-D3> Can freenet work through a socks proxy?
[15:33] <toad_> aaaargh
[15:33] <toad_> why did it work on the localtestnet?
[15:34] <Diablo-D3> I came up with a very odd idea that might be fun to try out
[15:34] <Diablo-D3> running freenet over tor.
[15:34] <iip_i2p> <lonelynerd> Diablo-D3: it might work with tsocks
[15:34] <toad_> possibly because I made lots of requests on the testnet...
[15:34] <Diablo-D3> iip_i2p is a relay bot?
[15:34] <toad_> yes
[15:34] <toad_> from IIP or is it I2P
[15:34] <Diablo-D3> wheres the other half?
[15:35] <Diablo-D3> heh
[15:35] <Diablo-D3> crap
[15:35] <Diablo-D3> debian doesnt package freenet
[15:36] <toad_> hmmm
[15:36] <toad_> why did it fall?
[15:36] <toad_> it was at 667% for several minutes
[15:36] <toad_> now it's 227%
[15:36] <Diablo-D3> freenet over tor should be a riot
[15:36] <iip_i2p> <lonelynerd> toad: tricky question, both iip and i2p users can access the same network nowadays
[15:36] <Diablo-D3> what is iip and i2p?
[15:37] <toad_> hmmm
[15:37] <toad_> the answer appears to be... that it received a request!
[15:37] <iip_i2p> <mule> toad_, incoming is based on average and controlled by requests sent? since i still see over 30k incoming with the limit set to 16k. of course with averages it would be a rather slow control loop, which could result in high input over many seconds or a few minutes.
[15:37] <Diablo-D3> hrm.
[15:37] <toad_> propagation load is outgoing reqs / incoming reqs
[15:37] <toad_> so if incoming reqs goes up, propagation load goes down...
[15:37] <toad_> hmmm
[15:37] <Diablo-D3> is there a webiste that describes how freenet works?
[15:38] <toad_> aha
[15:38] <iip_i2p> <mule> which would render output bw control rather useless to control link limits.
[15:38] <toad_> it does seem to be improving...
[15:38] <toad_> Diablo-D3: IIP and I2P are other anonymity projects which provide anonymous IRC
[15:38] <iip_i2p> <lonelynerd> Diablo-D3: that's probably because it's not so trivial to get working with gcj
[15:38] <Diablo-D3> toad_: ahh. I can do that through tor, actually.
[15:38] <toad_> 3 of the 4 nodes now have <600k MRIs
[15:38] <toad_> Diablo-D3: indeed
[15:38] <Diablo-D3> watch...
[15:39] <toad_> I haven't asked jr what the difference is recently
[15:39] * DiabloD3 (~diablo@puma.noncombatant.org) has joined #freenet
[15:39] <toad_> KenMan: verl: leex: make some requests please?
[15:39] <Diablo-D3> And...
[15:39] <iip_i2p> <lonelynerd> with tor you can use the webchat safely too, www.invisiblechat.com and get to this other side ;)
[15:39] <Diablo-D3> tada.
[15:39] <iip_i2p> <mule> so it would fulfill the purpose of overall traffic limit, which is nice for those hosts limited by monthly volume.
[15:40] <Kyon> me as well? :P
[15:40] <toad_> yeah
[15:40] <leex> i keep opn getting backe dof node
[15:40] * Diablo_D3 (~diablo@puma.noncombatant.org) has joined #freenet
[15:40] <Diablo-D3> oh bleh.
[15:40] <Diablo-D3> That wasnt supposed to happen.
[15:40] <Kyon> yeah have been backed off so far entirely
[15:40] <Diablo-D3> It was supposed to be another host =/
[15:41] * Diablo_D3 (~diablo@puma.noncombatant.org) Quit (Client Quit)
[15:41] * DiabloD3 (~diablo@puma.noncombatant.org) Quit (Client Quit)
[15:41] <toad_> KenMan: I think I'm measuring it the wrong way
[15:41] <toad_> KenMan: I'm doing load due to propagation = outgoing requests / incoming requests
[15:41] <toad_> this has some major bootstrapping problems
[15:41] <toad_> KenMan: any other ideas?
[15:41] <Diablo-D3> So, should I try freenet over tor?
[15:42] <iip_i2p> <mule> diablo-d3, why use tor? to hide that you use freenet?
[15:42] <toad_> hehe, who disconnected?
[15:42] <Diablo-D3> iip_i2p: just to see if it works, actually
[15:43] <Diablo-D3> iip_i2p: an anonymous connection to an anonymous network
[15:43] <toad_> seems little point in hiding the fact you use freenet by using another obvious anonymous net :)
[15:43] <Kyon> harder to track i suppose
[15:43] <Diablo-D3> erk, that should me mule:
[15:43] <Diablo-D3> damn you relay bot! damn yoooouuu!
[15:43] <toad_> somebody disconnected... and it was probably KenMan...
[15:43] <Kyon> toad_: still backed off
[15:44] <verl> up to 10 again
[15:44] <iip_i2p> <mule> diablo-d3: come over: www.invisiblechat.com or www.i2p.net
[15:44] <toad_> if that somebody decided "this sucks, I'm doing something else".. that's fair enough but could you TELL ME in future?!
[15:44] <Diablo-D3> mule: naw
[15:45] <toad_> verl: 10 what?
[15:45] <verl> Number of known routing nodes 12
[15:45] <toad_> this is absurd
[15:45] <toad_> the node starts up, # outgoing queries goes up a bit because of startup requests
[15:45] <toad_> # outgoing queries is low
[15:45] <toad_> so load due to propagation goes high
[15:45] <toad_> so we don't get any queries !!
[15:46] <verl> in 50 secs you are no longer backed off :)
[15:47] <toad_> Load due to propagation = 40.1% = 11.801675404683841 / 29.443276709550254
[15:47] <toad_> aha
[15:48] * Diablo-D3 downloads freenet-latest.tgz THROUGH TOR!
[15:48] * toad_ hmmm
[15:48] <toad_> possible fixes...
[15:48] <Kyon> i've recieved the cow files :-)
[15:48] <Diablo-D3> this is odd
[15:48] <toad_> ignore the first 20 output queries perhaps...
[15:48] <Diablo-D3> this is really odd
[15:49] <Diablo-D3> Im getting better performance through tor than I am through my normal connection
[15:50] <leex> if i am going from stable to unstable do i need to chage any thing in the freenet ini file
[15:51] <toad_> leex: no
[15:51] <toad_> probably
[15:51] <toad_> the bootstrapping problem, in detail:
[15:52] <toad_> on startup, load due to propagation goes high
[15:52] <toad_> it then won't recover for aaaaages, because everyone got sent a 10 minute MRI
[15:52] <toad_> so incoming reqs won't go high
[15:53] <toad_> solution: don't use the load until incoming reqs > some value
[15:54] <toad_> don't use the propagation load until incoming reqs received > some value
[15:55] <toad_> hmmm
[15:55] <toad_> now it's gone high again...
[15:57] <leex> woho i got an inbound request
[16:00] <KenMan> I still have a toadnode, it has drifted down to 24872ms MRI
[16:01] <leex> toad_ i got you 2x in Network Load
[16:02] <KenMan> If we want to generate requests, run this UNIX script - http://mywebpages.comcast.net/jkcorson/5089/bangOn5089.sh
[16:02] <toad_> hmmm
[16:02] <leex> as i have you as in incommming and as an out going cnnection
[16:02] <leex> so its has you in Network Load 2 times
[16:02] <toad_> both the incoming reqs and outgoing reqs seem to be absurdly variable...
[16:03] <toad_> the averages, that is
[16:03] <leex> shouldent it just be 1 in there
[16:03] <toad_> not the actuality
[16:03] <leex> as its counting you as 2
[16:03] <toad_> varies by a factor of 2 in a few seconds...
[16:04] <leex> ok
[16:05] <leex> just me not understanding it i think
[16:05] <KenMan> leex - it is probably an old reference for toad or something... pretty weird though
[16:05] <leex> its an fresh install
[16:05] <leex> and i know its you
[16:05] <leex> same key
[16:06] <leex> 0x0 : 6e4169afb67cc126112ac5a669c8c39eaa2ca274
[16:06] <leex> in both box's
[16:06] <KenMan> my node fingerprint on toadnode is 7c481b9299db253472ddbb318bd1ee4077adbd21
[16:06] <leex> how do you do an print scren
[16:06] <leex> hmm
[16:06] <leex> i only have 1 IP
[16:06] <toad_> KenMan: I think there are some problems in my new load estimation...
[16:07] <leex> from an VNC
[16:07] <leex> box
[16:07] * Kyon watches
[16:07] <KenMan> toad, i went looking for something that said 'give up on RNF after three outgoing attempts' but i couldn't find it anymore :(
[16:08] <toad_> hmmm
[16:09] * toad_ restarts node with bootstrapping and logging changes
[16:09] <toad_> done
[16:09] <KenMan> if you have Q multiplication much beyond 1.0000001 then the new load model may bring the network down, if it is trying to match in's to out's ...
[16:09] <toad_> KenMan: no, because it doesn't multiply queries
[16:10] <toad_> because of htl
[16:10] <KenMan> maybe. I am really not adding anything here, so I'll shut up :)
[16:10] <toad_> these averages change absurdly fast...
[16:10] <toad_> there must be some dumb bug somewhere...
[16:10] <KenMan> you aren't getting it then, if you think that reducing HTL on RNF avoids multiplication...
[16:10] <toad_> KenMan: of course there is multiplication
[16:10] <toad_> but it's limited
[16:11] <KenMan> the juggernaut is backoff
[16:12] <toad_> okay, now THIS is strange:
[16:12] <toad_> Reported 34456.0 on freenet.node.rt.TimeDecayingRunningAverage@152c4d
[16:12] <toad_> 9: weightedTotal=34456.00000135332, totalWeights=1.0000000000764968
[16:12] <toad_> Reported 311.0 on freenet.node.rt.TimeDecayingRunningAverage@152c4d9:
[16:12] <toad_> weightedTotal=28085.39004301258, totalWeights=1.8060828315546305
[16:12] <toad_> Reported 40794.0 on freenet.node.rt.TimeDecayingRunningAverage@152c4d
[16:12] <toad_> 9: weightedTotal=40794.000000014734, totalWeights=1.0000000000009475
[16:12] <toad_> it went up, then it went BACK DOWN !!
[16:13] <toad_> definitely a bug in TDRA
[16:13] <KenMan> :) toad likes to eat bugs
[16:14] <KenMan> but sometimes he doesn't thoroughly digest them, before excretion :o
[16:15] <toad_> KenMan: well, I'm trying to find out whether it's possible to do this without making RNFs fatal
[16:18] <toad_> okay
[16:18] <toad_> it's not a bug in TDRA
[16:18] <toad_> hmmm
[16:19] <KenMan> making RNFs fatal may kill some looped queries unfairly, but on a network like stable, that should only be a small percent of queries. I mean, look, 50% of queries manage to get something other than RNF, no ? In this highly congested environment...
[16:19] <leex> i can only one node
[16:19] <toad_> aha
[16:19] <KenMan> i am very confident that 90%+ of RNFs on stable are NOT due to looping. Rather it is localized congestion due to RNF sprawl.
[16:19] <toad_> halflife is way too low
[16:20] <toad_> halflife is in fact 1 second :)
[16:20] <KenMan> :)
[16:21] <KenMan> thank goodness it wasn't a large bug with a thick exoskeleton !! this one will go down easy...
[16:22] <toad_> okay, fixed it
[16:23] <toad_> anyone want new jar?
[16:23] <KenMan> insert the fixed jar through the network
[16:23] * jay (jcl@ool-18bf6dac.dyn.optonline.net) Quit (".")
[16:24] <KenMan> is there still a fixed upper limit on the number of retries ? it used to be 3 once upon a time...
[16:24] * ninja_ (im10ninjas@82-133-66-19.dyn.gotadsl.co.uk) Quit ("leaving")
[16:25] * ninja_ (~im10ninja@3ffe:bc0:8000:0:0:0:0:3fab) has joined #freenet
[16:25] <Kyon> mm sure
[16:25] * verl wants
[16:26] * toad_ restarts again :)
[16:27] <toad_> CHK@G34c9JW4g6wdaRplfz4N1Faur8YLAwI,AbbwV4Nyf-vo0piwea6PAA/freenet.jar
[16:27] <toad_> should call itself 60187
[16:28] * toad_ needs some queries...
[16:28] <verl> 5634 seconds more backoff
[16:28] <KenMan> verl it will drop before your eyes
[16:28] <verl> ah, yes
[16:28] <KenMan> 1881 seconds here
[16:28] * KenMan can be patient :)
[16:29] <toad_> Load due to expected outbound transfers: 219.5% because: 1000.0 req/hr 0.1(0 0s, 0 1s, 0 total) (pTransfer) 553225.0 bytes = 55322500 bytes/hr expected from current requests, but maxOutputBytes/minute = 420000 * 60 = 25200000 bytes/hr target
[16:29] <toad_> look at this!
[16:29] <toad_> argh
[16:29] <KenMan> argh indeed
[16:29] <verl> 1900 secs...you are ahead
[16:30] <toad_> okay, new jar... :)
[16:30] <toad_> KenMan: the point is, 1000.0 is the default!
[16:30] * toad_ makes that not count until it has at least 10 requests
[16:31] * greycat (~wooledg@192.35.79.70) Quit ("Lick Bush in '04")
[16:31] <Kyon> hmm, how can you see how many secs?
[16:31] * remote-lex (~le@spc1-warr1-3-0-cust198.bagu.broadband.ntl.com) has joined #freenet
[16:31] <remote-lex> lo agane
[16:32] <verl> Kyon: go to Node Status Interface
[16:32] <remote-lex> time to start dcc my self
[16:32] <verl> then Node Reference Status
[16:32] <toad_> load is 10%...
[16:32] <toad_> but outgoing MRI is 600k
[16:32] <verl> then scroll down... to the table
[16:32] <toad_> because we have had too many outgoing reqs...
[16:32] <toad_> or something
[16:33] <toad_> some perversity of rate limiting...
[16:33] <verl> scroll right, and look at Minimum Request Interval
[16:33] <Kyon> ahh
[16:33] <Kyon> 656 secs
[16:33] <Kyon> gotcha
[16:33] <toad_> ahh
[16:33] <toad_> better
[16:34] <toad_> now it's 110 secs
[16:34] <toad_> well 9 secs to 240 secs actually :|
[16:34] <verl> 12244ms :)
[16:34] <toad_> still no requests though...
[16:34] <toad_> 12 secs
[16:35] <toad_> still no requests..
[16:35] <toad_> verl: make some requests :)
[16:35] <verl> backed off !
[16:35] <toad_> not for long surely? my outgoing MRI is only 8 seconds
[16:35] <verl> well now i am downloading
[16:36] <toad_> yay, got some
[16:36] <toad_> ok
[16:36] <Kyon> i have it
[16:36] <toad_> okay, try this one:
[16:36] <Kyon> waiting..
[16:36] <toad_> (the last one wasn't as good)
[16:37] <toad_> CHK@Csyh7kfmhxW6E7GhDdPyVNjsCF0LAwI,HcDtsA7hIrRbt-sTN-EZIA/freenet.jar
[16:37] <toad_> Load due to propagation = 268,865,340.2% = 113.6481732515818 / 4.226955143235222E-5
[16:37] <toad_> aieeeeeee
[16:38] <toad_> Reported 1.092084782882E12 on freenet.node.rt.TimeDecayingRunningAve
[16:38] <toad_> rage@111a775:
[16:38] <toad_> duhh :)
[16:40] * KenMan still doesn't understand how a CHK can be a splitfile (exceed 1MB), but right now he doesn't need to
[16:40] <toad_> okay, new jar soon...
[16:41] <toad_> KenMan: you can have redirects which are also CHKs
[16:41] <KenMan> i guess as much, i just don't know much about redirects...
[16:42] <KenMan> is anyone around that is running stable ??
[16:43] <KenMan> capping maximum retries to one on stable might be an incremental improvement, one that accomodates a single loop per hop anyway.
[16:44] <KenMan> in that case, searchFailedCount would never be allowed to exceed one.
[16:45] <toad_> okay, new jar:
[16:45] <toad_> CHK@kRNAMtlh0Mjf~zmQcNLxs9h~h38LAwI,knCoveUCKWscYhbYAxQPqg/freenet.jar
[16:45] * mikeDOTd (~mikedotd@plutonium.nexthop.net) has joined #freenet
[16:46] <toad_> bbiab
[16:46] <toad_> going to get food
[16:46] <toad_> suggest you get the jar and restart
[16:46] <KenMan> how does MRI=600627 translate into backed off until 2525 seconds ??
[16:46] <KenMan> food good
[16:46] <KenMan> hunger bad
[16:46] <toad_> KenMan: it has to make the average as well as the intervla
[16:46] <KenMan> ah...
[16:47] <toad_> bbl
[16:47] <toad_> tell me how it works out
[16:47] <KenMan> will do
[16:49] <KenMan> verl kyon - anybody have more than one route yet ?
[16:49] <Kyon> only 1
[16:49] <Kyon> backed off still
[16:49] <KenMan> I'm backed off for another 40 minutes... bb30
[16:51] <Kyon> yeah i have 30 secs
[16:51] <Kyon> err
[16:51] <Kyon> 30 mins
[16:51] <Kyon> i need to run as well, not sure when ill be back
[16:57] <verl> it is not possible to download this file
[16:57] <remote-lex> dam this pc got ad crap on it it will not let me open loacl host
[16:57] <verl> it starts the download and keeps it up for 30 secs, same every time
[16:57] <verl> then it fails without one block complete
[17:04] * remote-lex (~le@spc1-warr1-3-0-cust198.bagu.broadband.ntl.com) Quit (Read error: 104 (Connection reset by peer))
[17:18] <KenMan> ah hah, had to compensate for rate limiting on my 15th try... now it is coming across
[17:18] <KenMan> set simultaneous downloads to 1
[17:20] <KenMan> got 6 of 10
[17:20] <verl> i got it!
[17:20] <KenMan> why does this feel like a game of chess ? I don't really care for chess...
[17:20] <verl> i did work when i set it at 1 simultaious
[17:21] <KenMan> guess we have to outthink the rabbit...
[17:22] <KenMan> failed at 8 of 10... i'm getting there, with lots of effort :(
[17:22] * cehteh (foobar@cehteh.homeunix.org) has left #freenet
[17:25] * remote-lex (~le@spc1-warr1-3-0-cust198.bagu.broadband.ntl.com) has joined #freenet
[17:25] <remote-lex> back
[17:26] <remote-lex> any one want to fri an egg
[17:26] <remote-lex> got my stove (CPU) at 67 here
[17:26] * remote-lex TEMP
[17:27] * KenMan is on his 7th splitfile attempt, with 9 of 10 blocks received. Check but no checkmate.
[17:29] <remote-lex> i cant get 1 block
[17:31] <KenMan> finally, i got the new jar.
[17:31] <verl> next time im putting it up on the web...
[17:32] <leex> download fnished
[17:33] <verl> hey, got some traffic now
[17:33] <verl> Connections open (Inbound/Outbound/Limit) 3 (1/2/200)
[17:35] <remote-lex> probly me
[17:37] <remote-lex> seems to be working on this 10 min old node
[17:37] <remote-lex> :)
[17:38] <KenMan> i have 3 peers 82.32. 213.106. 217.208.
[17:38] <remote-lex> 1 is me
[17:38] <verl> i also have a guy namned void!
[17:39] <remote-lex> yer
[17:39] <KenMan> really ! do you have me ? 68.34.
[17:39] <remote-lex> its an none IP guy
[17:39] <remote-lex> yep
[17:39] <KenMan> let's figure out who it is then. There's no anonymity here...
[17:40] <KenMan> verl list your peers you do have IPs for, please.
[17:40] <remote-lex> its what freenet does probly an ip that has droped out but the key has been kept
[17:40] <remote-lex> so the ip is voide
[17:40] <Diablo-D3> okay
[17:40] <Diablo-D3> now lets see
[17:40] <Diablo-D3> freenet over tor
[17:41] <remote-lex> you should see 2x 82. ips
[17:41] <verl> i have for 82.32 213.106 and 68.34
[17:42] <verl> altohugh 213 guys has not yet upgraded and is downloading the jar from me, i belive
[17:42] <remote-lex> ops
[17:42] <Diablo-D3> ...freeenet over toooor
[17:43] <remote-lex> 1 82 and 1 213ip
[17:43] <remote-lex> probpy me
[17:43] * Diablo-D3 rushes up to the edge of the cliff, and slowly backs away
[17:43] <remote-lex> lol
[17:43] <Diablo-D3> freenet over tor.
[17:43] <verl> now here comes mr 141!!
[17:43] * Diablo-D3 rushes up to the edge of the cliff, and slowly backs away
[17:43] <leex> any thing els to test
[17:43] <Diablo-D3> freenet over tor!!! WAAAASAARRTRTGGGGHHH!!!!!!
[17:43] * Diablo-D3 jumps over the cliff
[17:44] <KenMan> sounds like a downhill path to me ...
[17:44] <verl> we must wait for toad...
[17:44] <verl> he will know
[17:44] <Diablo-D3> shit
[17:44] <remote-lex> how that work
[17:44] <KenMan> did gravity catch up with you ??
[17:45] <Diablo-D3> lets ln -s ./usr/lib/j2re1.5-sun/bin/java /usr/bin/java first
[17:45] <remote-lex> tor over freenet
[17:45] <Diablo-D3> so how do I know if this is working?
[17:45] <KenMan> freenet and tor in gravitational equilibrium ... woah.
[17:46] <Diablo-D3> Im assuming I can just torify ./start-freenet.sh right?
[17:46] <KenMan> tor and freenet, between the sheets !! wahhh AAHHHH...
[17:46] <remote-lex> lol 2x ammoty
[17:46] <remote-lex> well cida
[17:47] <remote-lex> not so good for the toor user who requested for the web site
[17:47] <remote-lex> to forwd it to you
[17:47] <remote-lex> forwad
[17:47] <Diablo-D3> er
[17:47] <Diablo-D3> what is the fred port again?
[17:47] <KenMan> 8888 ?
[17:47] <remote-lex> yep
[17:47] <Diablo-D3> 1 0:32.49 0.0 66.7 35 10 S 328m 182m 145m 1000 java
[17:47] <Diablo-D3> eek!
[17:48] <KenMan> oh, look for listenPort if you want FNP
[17:48] <Diablo-D3> virts 328, swap 181, res 146!
[17:48] <Diablo-D3> thats insane!
[17:48] <remote-lex> freenet downloaded
[17:48] <remote-lex> test file
[17:49] <KenMan> I'm going to start Frost on this network :o
[17:50] <Diablo-D3> hrm
[17:50] <Diablo-D3> what ports do I have to open for freenet?
[17:50] <mazzanet> the one you specifiy
[17:50] <mazzanet> -i
[17:50] <Diablo-D3> whats the default?
[17:50] <mazzanet> its random
[17:50] <remote-lex> randome
[17:50] <mazzanet> check freenet.conf/ini
[17:51] <remote-lex> look in the freenet ini file
[17:51] <remote-lex> heh
[17:51] <KenMan> rahndahm
[17:51] <mazzanet> it'll be "listenPort"
[17:51] <remote-lex> its randome for fist time run
[17:51] <Diablo-D3> heh
[17:51] * Diablo-D3 will have to fix that
[17:51] <KenMan> ooh, i have received up to 11 queries/min on this network.
[17:51] <remote-lex> it stays as long as you not a chage it or b you del the freenet.ini file
[17:53] <remote-lex> doadoa
[17:54] <remote-lex> grate when things work
[17:58] <remote-lex> any one want to see this
[17:58] <remote-lex> jpg
[17:59] <remote-lex> when toad_ back
[18:00] <remote-lex> any way me needs an reboot as i have broke something on this pc
[18:00] <remote-lex> like nothing is starting up with windows
[18:00] * remote-lex (~le@spc1-warr1-3-0-cust198.bagu.broadband.ntl.com) Quit ()
[18:05] * remote-lex (~le@spc1-warr1-3-0-cust198.bagu.broadband.ntl.com) has joined #freenet
[18:05] <Diablo-D3> well that was short lived
[18:06] <toad_> hi
[18:07] <toad_> Load due to propagation = 70.9% = 88.87677633598388 / 125.34866911683785
[18:07] <toad_> anyone here? leex ? KenMan ? verl ?
[18:07] <toad_> Kyon ?
[18:07] <KenMan> back
[18:08] <remote-lex> toad_
[18:08] <remote-lex> plus my home pc
[18:08] <remote-lex> seems to be working grate now
[18:08] <KenMan> are you getting any queries from my frosted node ?
[18:10] <remote-lex> how i know that
[18:10] <toad_> KenMan: i think somebody is - 2 nodes have 600k+ mri's
[18:10] <KenMan> should i stop then ?
[18:10] * toad_ is getting quite a few queries total - 2-4/min usually
[18:10] <toad_> which IP is which?
[18:10] <KenMan> im 68.34.
[18:11] <toad_> i'm 82.32.16.91
[18:11] <toad_> KenMan: did you upgrade?
[18:11] <verl> 217.208
[18:11] <toad_> your MRI is crazy
[18:11] <toad_> is that because you are running Frost? :)
[18:11] <KenMan> yeah, i upgraded
[18:11] <toad_> okay
[18:11] <toad_> go to Environment
[18:12] <toad_> do you have a lot of threads running? what is your global quota (General) ?
[18:12] <remote-lex> lex is 82.0.
[18:12] <toad_> also what is your load due to propagation, remote-lex and KenMan?
[18:12] * Diablo-D3 decides to upgrade his java
[18:12] <remote-lex> and i am 213.106.
[18:12] <toad_> remote-lex: that's the other overloaded one
[18:12] <toad_> did it get upgraded?
[18:12] <toad_> evidently not
[18:12] <toad_> that's why
[18:13] <Diablo-D3> 06/30/03 - If you are using Internet Explorer to download the Linux Java plug-in, use right-click to save this file to your computer. Use 'Save Target As'.
[18:13] <remote-lex> ?
[18:13] <toad_> so 213.106... is in high MRI because it's using the old version
[18:13] <toad_> and kenman's is because it's running frost
[18:13] <KenMan> see , even a 4 node network cannot keep up with toad's rapid pace of upgrades ;)
[18:13] <toad_> KenMan: apparently not :|
[18:13] <Diablo-D3> -Java.com Consumer Support
[18:13] <remote-lex> lol
[18:14] <toad_> KenMan: could you get me the stats i asked for?
[18:14] <remote-lex> that what the freenet.jar file was for
[18:14] <KenMan> Pooled threads in use 3 of 27
[18:14] <Diablo-D3> Am I the only one here who sees how _wrong_ that is?
[18:14] <toad_> 3 of 27?!
[18:14] <toad_> with Frost running?
[18:14] <Diablo-D3> 27!?
[18:14] <toad_> then why is the MRI so high?
[18:14] <KenMan> current estimated load is 6494% ...
[18:14] <toad_> what's your globalQuota?
[18:14] <toad_> ouch
[18:14] <toad_> why?
[18:14] <remote-lex> freenet.jar < the one we downloaded
[18:15] <remote-lex> we use now
[18:15] <KenMan> current Global Quota = 533.11
[18:15] <remote-lex> useing freenet
[18:15] <toad_> this one?: CHK@kRNAMtlh0Mjf~zmQcNLxs9h~h38LAwI,knCoveUCKWscYhbYAxQPqg
[18:15] <toad_> KenMan: reasons for load?
[18:15] <toad_> Current estimated load for QueryReject purposes 3%
[18:15] <toad_> Current estimated load for rate limiting 95.4%
[18:15] <toad_> Load due to propagation = 95.4% = 114.70612408462907 / 120.20112003734941
[18:15] <KenMan> load due to propagation
[18:15] <toad_> KenMan: really? hrrrm
[18:16] <toad_> that's strange
[18:16] <verl> Current estimated load for rate limiting 10%
[18:16] <KenMan> I AM running frost , after all.
[18:16] <toad_> KenMan: ahhh
[18:16] <toad_> all the outgoing queries
[18:16] <toad_> ok
[18:16] <KenMan> shall i stop it ?
[18:16] <toad_> KenMan: any way to tell how many queries it is responsible for?
[18:16] <KenMan> yes, hold on
[18:17] <KenMan> localRequestSuccessRatio says about 80/minute
[18:17] <toad_> woah
[18:17] <toad_> 80/min ?!
[18:17] <toad_> woah
[18:17] <KenMan> like 75 to 75
[18:17] <toad_> why?
[18:17] <KenMan> to 85
[18:17] <toad_> downloading or just boards?
[18:17] <toad_> and for how long is this sustained?
[18:17] <KenMan> it is just scanning for messages on several hundred boards
[18:18] <toad_> why do you have several hundred boards?
[18:18] <KenMan> i been doing 80/min for about 25 minutes
[18:18] <toad_> woah
[18:18] <toad_> and since that's the only source of load on the network... your propagation load is really high
[18:18] <toad_> okay
[18:18] <KenMan> not that it makes any difference but i haven't run frost for a long time
[18:18] <toad_> how can we have our cake and eat it too?
[18:19] <toad_> for instance... what would happen if we ignored locally originated requests for load due to propagation?
[18:19] <remote-lex> upgraded now
[18:19] <toad_> a single attacker should not be able to tell how many requests we are making in total
[18:19] <remote-lex> do i need an new unstable.ref
[18:19] <toad_> remote-lex: no
[18:19] <toad_> just the same 4 node testnet :)
[18:20] <toad_> KenMan: any other obvious possibilites?
[18:20] <remote-lex> i did not know it was an new ver heh
[18:20] <KenMan> your rubber-band solution is an easy path to try (ignore local reqs for load prop)
[18:20] <KenMan> but it will not prevent people from running frost.
[18:21] <toad_> well yes
[18:21] <toad_> so?
[18:21] <toad_> the load will propagate back to the node anyway
[18:21] <KenMan> it will cause waves of congestion on nodes closest to the frost user
[18:21] <toad_> well lets see
[18:21] <toad_> the node with the frost user happily sends out the maximum allowed requests
[18:21] <KenMan> and takes a dump on all its peers
[18:21] <toad_> nearby nodes get a little overloaded, and reduce
[18:22] <toad_> but not by too much since frost requests don't actually cause all that much load
[18:22] <toad_> KenMan: well, the alternative is pretty much unthinkable
[18:22] <KenMan> so don't think it, whatever it is
[18:22] <toad_> if we only allow outgoing requests equal to the number of requests we terminate, we'd only allow a VERY small number of requests
[18:23] <KenMan> frost doubles load. It guarantees that the user's node will reject almost all external queries
[18:23] <Diablo-D3> what version of java are you guys using?
[18:23] <toad_> of course we could try to get a decent background load for a better simulation...
[18:23] <toad_> KenMan: REJECT ?!?!
[18:23] <KenMan> in favor of the frost ones...
[18:23] <KenMan> yes, RNF due to backoff. REJECT
[18:23] <toad_> KenMan: surely it'll just send out crazy MRIs, not actually QR?
[18:23] <verl> Diablo-D3: 1.4.2_05-b04 works fine fore me
[18:24] <toad_> KenMan: ahhhh
[18:24] <toad_> hmmm
[18:24] <KenMan> no, it will QR because of nonexistent routes
[18:24] <Diablo-D3> so the newest 1.4.x should work?
[18:24] <KenMan> QR(RNF) that is
[18:24] <toad_> okay, what's the solution to that?
[18:24] <KenMan> heh, i won't fall for your trap today
[18:24] <verl> Diablo-D3: yes, 1.4.1 may have issues, dont remember
[18:25] <toad_> I'm not sure that we CAN queue FCP requests
[18:25] <toad_> because the client will just reissue them
[18:25] <KenMan> i know, but i think the lightbulb just flashed in your brain...
[18:25] <Diablo-D3> well, 1.5 has _serious_ issues
[18:25] <toad_> IMHO we probably want to tell frost how often it can issue requests
[18:25] <toad_> Diablo-D3: it does?
[18:25] <KenMan> yeah. same problem on splitfiles (apparently)
[18:25] <toad_> verl: some vers of 1.4.2 have big problems
[18:25] * kers (~kers@192.ppp141.rsd.worldonline.se) Quit ("Leaving")
[18:25] <KenMan> ( simultaneous downloads setting )
[18:25] <toad_> not the most recent one though
[18:26] <Diablo-D3> toad_: yeah, eats a shitload of memory
[18:26] <toad_> KenMan: well that's not the only way to deal with it...
[18:26] <verl> toad: all jvms have issues depending on who you talk to
[18:26] <toad_> KenMan: so how do we do it? Give each and every client a requestinterval every so often?
[18:26] <toad_> verl: indeed :<
[18:26] <KenMan> that is an option, i suppose.
[18:27] <toad_> KenMan: unfortunately that means we have to know which conns belong to which client
[18:27] <KenMan> i'm glad you see the purpose for needing it though.
[18:27] <toad_> KenMan: well we can't just queue them, they'll assume they're hung and just retry
[18:27] <KenMan> which is fine, just update the entry in the queue
[18:27] <Diablo-D3> man, 14 megs is huge
[18:27] <toad_> alternatively, we could simply expect the client authors to reduce the request frequency if every request is RNFing :)
[18:28] <KenMan> assign it to the new connection instead of the old one
[18:28] <toad_> KenMan: what if they retry with a different request?
[18:28] * Diablo-D3 download j2re-1.4.2-05
[18:28] <toad_> e.g. in the case of a splitfile download?
[18:28] <KenMan> they get queued
[18:28] <toad_> perhaps a client-cache would be useful here...
[18:28] <verl> toad: you can put it in the text when one gets a rnf "dont hammer the retry button!!"
[18:28] <KenMan> i really don't want to make a lot of suggestions to you, but you see the problem now, right ?
[18:28] <toad_> verl: yeah, but I'm talking application level here...
[18:29] <verl> and if he continues have it crash the webbrowser or something
[18:29] <toad_> lol
[18:29] <Diablo-D3> man this is going to take me like a half hour
[18:29] <toad_> KenMan: yeah
[18:29] <KenMan> then i am content
[18:29] <toad_> although i think for things like Frost, it'd make a big difference having fixed key sizes
[18:29] <toad_> frost only needs to fetch a certain number of smallish keys to check the boards
[18:29] <Diablo-D3> what is frost?
[18:30] <verl> Diablo-D3: jtcfrost.sourceforge.net
[18:30] <KenMan> a spamming tool
[18:30] <toad_> these get counted as full size requests - most of which are 1MB
[18:30] <toad_> lol KenMan
[18:30] <Diablo-D3> heh
[18:30] <toad_> Diablo-D3: message boards over freenet
[18:30] <Diablo-D3> my bandwith is in use
[18:30] <Diablo-D3> toad_: ahh cool
[18:30] <toad_> okay, here's an idea:
[18:30] <toad_> implement a new freenet client protocol
[18:30] <Diablo-D3> what I really want is web browsing over freenet
[18:31] <toad_> somewhat similar to the old one, except that it multiplexes, and returns MRI, and exposes the queueing structure, and allows status queries for running or queued requests
[18:31] <KenMan> toad, do you see that the focus of the problem occurs in the set of backoff values, and degree of backoff ?
[18:31] <toad_> Diablo-D3: you can browse the free-web :)
[18:31] <Diablo-D3> toad_: ?
[18:31] <Diablo-D3> actually, that would be fucking cool
[18:31] <toad_> Diablo-D3: there is a lot of web-like content on freenet
[18:32] <toad_> you just install it and click on the links
[18:32] <Diablo-D3> a http proxy that connects to freenet
[18:32] <toad_> you just can't get to the Real Web
[18:32] <KenMan> interarrival times of queries, versus instants when routes become available ?
[18:32] <Diablo-D3> toad_: no no no
[18:32] <toad_> KenMan: hmmm?
[18:32] <Diablo-D3> toad_: you embed a tag in your website that gives a freenet key
[18:33] <Diablo-D3> toad_: then the proxy picks this up, and grabs the freenet version instead
[18:33] <KenMan> the crux of your whole rate limiting troubles has to do with the patterns of routes going backed off and becoming available... and the patterns of query arrivals...
[18:33] <toad_> Diablo-D3: but you've just fetched it...
[18:33] <Diablo-D3> toad_: no I havent
[18:33] <Diablo-D3> toad_: I just fetched a tag.
[18:34] <Diablo-D3> toad_: everything would be on freenet, images, links, _everything_
[18:34] <toad_> Diablo-D3: and thus blown your anonymity
[18:34] <KenMan> the composite (for all requestors) interarrival times for queries is a quasi-random function
[18:34] <toad_> KenMan: perhaps
[18:34] <Diablo-D3> toad_: feh
[18:34] <toad_> if you can keep % backed off down to 25 or so, then it shouldn't be such a huge problem
[18:34] <KenMan> you cannot synchronize it with routes becoming available
[18:34] <Diablo-D3> toad_: fetch it through tor ;)
[18:35] <KenMan> but you won't be able to do that without freezedown. This is why I proposed a method that would provide 50%.
[18:35] <toad_> KenMan: well, that function is largely a problem because the backoff times are stupendously high
[18:35] <toad_> KenMan: uhh, no
[18:35] <toad_> :)
[18:35] <KenMan> and, hand-in-hand, backoff percents are 90%+ at any instant
[18:35] <toad_> it wouldn't, it's just some statistical doublethink
[18:35] <toad_> right
[18:35] <KenMan> you have written it off without full understanding
[18:36] <toad_> but IF the code we have RIGHT NOW works on the larger network, then IMHO it should provide considerably better than the 90% or so that we get now
[18:36] <toad_> in fact I wouldn't be surprised if by itself it gave 50%
[18:36] <KenMan> let me tell you how to get 50% backoff using your current limiting scheme then.
[18:36] <remote-lex> lol
[18:36] <Diablo-D3> hrm
[18:36] <KenMan> aim for half the qph rate that any given mri equates to.
[18:36] <toad_> KenMan: by changing one line in StandardNodeEstimator so that it's the same as your strange png?
[18:37] <toad_> KenMan: that will definitely cause freezedown
[18:37] <Diablo-D3> heheh
[18:37] <Diablo-D3> there has to be a way to use tor!
[18:37] <KenMan> if a node says, backoff for 50ms, then aim to stay off him for 100ms
[18:37] <toad_> what do you mean by "aim" ?
[18:37] <toad_> what level would you implement this at?
[18:37] <Diablo-D3> who here thinks we need to do freenet over tor?
[18:37] <KenMan> construct your algorithms to target 50% use, not 100% (which equates to full backoff / exhaustion)
[18:37] <toad_> which algorithms?
[18:38] <KenMan> the ones that compute rates and produce MRIs
[18:38] <toad_> WHICH ONES?
[18:38] <toad_> I mean, do you mean the rate limiting stuff?
[18:38] <toad_> do you want me to just double the load-due-to-propagation perhaps?
[18:39] * toad_ is pretty sure that would result in meltdown but I suppose we could try it
[18:39] <KenMan> gah... okay, you like network fabric analogies. Mine is most closely related to ATM, which used small fixed packet sizes.
[18:39] <KenMan> Consider each MRI period as a cell.
[18:39] * toad_ has read a little about ATM
[18:39] <toad_> okay, so you divide time up into MRI periods
[18:39] <toad_> in each one, either there's a request or there isn't
[18:39] <toad_> right?
[18:40] <KenMan> in your model, you allow the cells to grow larger than a cell size. The cell size is reserved for "no queries". The amount of growth depends on utilization.
[18:40] <KenMan> yeah, that's my model
[18:40] <KenMan> which keeps every frame the SAME size.
[18:40] <toad_> whereas my model is to put the requests on the boundaries between cells?
[18:40] <KenMan> yes
[18:40] <toad_> i don't really see how they're different
[18:40] * Hory (~Miranda@81.196.25.110) Quit (Read error: 104 (Connection reset by peer))
[18:40] <toad_> if you're saturated, you're still going to send exactly one request every N millis, aren't you?
[18:41] <KenMan> and , your rate/Q limiting algorithm is aiming to minimize the space between the cells, which maximizes backoff (approximates exhaustion)
[18:41] <toad_> if you're not, you're going to wait until you've got something to send
[18:41] * Diablo-D3 sobs
[18:41] <Diablo-D3> no one wants freenet over tor =/
[18:41] <toad_> KenMan: it is?!
[18:41] <toad_> how, exactly? it certainly isn't a design goal
[18:41] <KenMan> but, the percent of time that any single route is available is being minimized, if you maximize your use.
[18:41] <remote-lex> toad_ going off topic
[18:42] <toad_> remote-lex: huh?
[18:42] <remote-lex> tor thingi
[18:42] <remote-lex> unless you wan it in
[18:42] <toad_> KenMan: yes, and we WILL maximize our usage, IF we need to
[18:42] <toad_> and as a practical reality, on good nodes, we do need to
[18:42] <KenMan> in my model, the requestor is free to select the exact moment when he can utilize a specific route. One that coincides with the quasi-random occurrence of interarrival times
[18:42] <toad_> remote-lex: i don't mind
[18:42] <remote-lex> :)
[18:42] <Diablo-D3> well, Im thinking this
[18:42] <toad_> KenMan: that is the case in my model too
[18:43] <Diablo-D3> I need to get the tor guys to make an api
[18:43] <toad_> KenMan: since we don't queue requests
[18:43] <KenMan> in your model, 100% utilization of a route means it is backed off 100% of the time, thus unavailable
[18:43] <toad_> KenMan: no
[18:43] <remote-lex> toad_ just another task to keep you thinking :))
[18:43] <KenMan> you use it the instant it becomes available, no ?
[18:43] <Diablo-D3> er, a library I mean
[18:43] <toad_> because it's highly unlikely there's a request instantly upon it becoming available
[18:43] <toad_> because we don't queue!
[18:43] <Diablo-D3> and we just tell freenet to use this library, and peddle in onion hosts instead
[18:43] <toad_> if we queued, then perhaps
[18:43] <toad_> Diablo-D3: why would you want to?
[18:44] <toad_> Tor doesn't provide any meaningful steganography, does it?
[18:44] <KenMan> i know that, i'm just asking a question. If MRI=100ms = 600 qph, that absolutely demands 100% usage, no available times.
[18:44] <toad_> and if Freenet is illegal,
[18:44] <Diablo-D3> toad_: anonymous connections to other free-tor users.
[18:44] <toad_> Tor surely would also be
[18:44] <KenMan> my math is bad, but you can get the idea
[18:44] <Diablo-D3> freenet and tor are both illegal in some country
[18:44] <toad_> KenMan: eh?
[18:44] <toad_> okay
[18:44] <toad_> if we have MRI = 100ms
[18:44] <Diablo-D3> face it, its a fact of life.
[18:44] <toad_> and we send 600 qpm
[18:44] <toad_> then we will have zero interarrival times
[18:45] <toad_> err
[18:45] <toad_> zero Tgap
[18:45] <KenMan> zero intercell gaps yes
[18:45] <toad_> = distance between when we CAN send and when we DO send
[18:45] <toad_> but in practice that WILL NEVER HAPPEN, until we implement queuing
[18:45] <toad_> Diablo-D3: indeed. Your point?
[18:45] <KenMan> it will get really damned close, which is why backoff is really damned bad
[18:46] <Diablo-D3> toad_: I dunno, I thought it would be cool
[18:46] <toad_> KenMan: if we are busy then under your model, we will get really damned close to sending a request the moment the new cell starts
[18:46] <KenMan> so, you will never be able to use the full capacity that you are using in your algorithms
[18:46] <toad_> KenMan: no, it's exactly the same
[18:46] <toad_> assuming heavy load, then Tgap -> 0
[18:46] <Diablo-D3> come on java! go go go!
[18:46] <KenMan> this is true, but I can still gaurantee 600qph in my model. And accomodate for less than perfect coincidental timings.
[18:47] <toad_> but in your model, Tgap is the difference between when you send it and when the next cell opens and you're first allowed to send it
[18:47] <toad_> that will ALSO go to 0
[18:47] <KenMan> for every loss of exact precision on your model, you are losing qph
[18:47] <Diablo-D3> so, guys
[18:47] <toad_> KenMan: what do you mean by less than perfect coincidental timings?
[18:47] <Diablo-D3> what can I use freenet for?
[18:47] <remote-lex> i could read you lot talking all night
[18:47] <toad_> I don't see that you can accomodate it actually
[18:47] <toad_> any more than I can
[18:47] <KenMan> interference caused by the collective state of routes "availability"
[18:47] <toad_> if we're busy we'll get the first request after we become available, either way
[18:48] <toad_> then we have to wait a long time till we're available again
[18:48] <toad_> Diablo-D3: all sorts of things
[18:48] <KenMan> no, im referring to reality, where queries don't arrive exactly every 600qph/whatever ms
[18:48] <toad_> Diablo-D3: sharing the Operating Thetan papers e.g.
[18:48] <toad_> KenMan: right, so?
[18:48] <Diablo-D3> heh
[18:48] <toad_> how does your model help with that?
[18:48] <KenMan> without queueing, you require nature to do something that it just won't do. In my model, I can do it anyway
[18:48] <toad_> queueing would help, of course
[18:49] <toad_> you can? how?
[18:49] <Diablo-D3> does anyone actually think the scientologists exist?
[18:49] <Diablo-D3> its really just a giant plot by the government
[18:49] <toad_> Diablo-D3: their legal department apparently does
[18:49] <toad_> Diablo-D3: there's lots of government conspiracy stuff too
[18:49] <toad_> freesites such as Thought Crime
[18:49] <toad_> and Bad Shit if True (debatable legality, lots of video)
[18:50] <toad_> and the Diebold files
[18:50] <Diablo-D3> well, soon as j2 1.4.2 is done downloading...
[18:50] <toad_> KenMan: okay...
[18:50] <toad_> okay I see
[18:50] <toad_> well it was an intentional safeguard I introduced at some point
[18:50] <KenMan> this stuff is so hard to explain. But I will not come down off my pedestal on this issue. You need to spend more time "looking" at what the collective backoff states are doing in millisecond timings...
[18:50] <toad_> both satisfy the average AND the individual MRI
[18:50] <Diablo-D3> so, hrm.
[18:50] <toad_> perhaps it's unnecessary
[18:51] <toad_> and yes, the way I do it now, WILL mean we can't get total 100% usage
[18:51] <Diablo-D3> what are you two arguing about?
[18:51] <toad_> otoh it's a safeguard - it's slack space
[18:51] <KenMan> math, i think.
[18:51] <Diablo-D3> well, hows this
[18:51] <toad_> I can take it out if you think it would make more than a percentage point difference in an hour
[18:51] <Diablo-D3> toad_: you do your thing
[18:51] <Diablo-D3> KenMan: you do your thing
[18:52] <Diablo-D3> and you both come back and see which one had better performance
[18:52] <toad_> Diablo-D3: sadly it's pretty hard to do that :)
[18:52] <KenMan> exactly
[18:52] <toad_> because we're dealing with network wide phenomena
[18:52] <Diablo-D3> why?
[18:52] <Diablo-D3> feh
[18:52] <KenMan> but we have fun charging at each other, anyway :)
[18:52] <Diablo-D3> then this problem probably has already been solved
[18:52] <remote-lex> nen
[18:52] <verl> why not make 2 testnets, one each?
[18:52] <toad_> KenMan: okay, so we'd get a slightly higher request rate for a given MRI
[18:52] <remote-lex> ?
[18:52] <toad_> otoh, we
[18:53] <toad_> we'd get a correspondingly greater incidence of MRI violations
[18:53] <Diablo-D3> so, hrm
[18:53] <toad_> however that may be wrong thinking - MRI violations are probably caused by bugs...
[18:53] <Diablo-D3> damnit, isnt tor handy somewhere?
[18:53] <KenMan> remember, we are doing this on both sides, ins and outs. Cells are deterministic.
[18:54] <KenMan> they have a perfect limit that is reachable, not just approachable.
[18:54] <toad_> KenMan: well, doing it in cells ~= checking the average (but not the individual cell timing)
[18:54] <toad_> I've explained above why i think it is useful
[18:54] * studdlybiggi (~noneya@12-215-171-38.client.mchsi.com) has joined #freenet
[18:54] <studdlybiggi> hello
[18:54] <toad_> I don't know whether that justifies having it
[18:54] <toad_> perhaps it doesn't
[18:54] <toad_> hi studdlybiggi
[18:54] <KenMan> it allows for some wiggliness that you are forced to deal with - random interarrival times
[18:55] <toad_> KenMan: however, I still don't see how it would cause 50% backoff
[18:55] <toad_> KenMan: it allows for slightly more wiggliness
[18:55] <toad_> we already have wiggliness forwards (after the availability)
[18:55] <toad_> it would distribute that backwards a bit too
[18:55] <KenMan> IF you equate cell length to time, and compute cell length based the way you do now, you can permit non-uniform (random) interarrivals
[18:56] <toad_> but as load increases, wiggliness (rapidly) decreases
[18:56] <KenMan> wiggliness for random instants, but still the guarantee of a hard limit, which is what you want
[18:56] <KenMan> no, it doesn't on a millisecond scale. You never reach enough load to have uniformity
[18:56] <toad_> KenMan: you agree that if I take out the absolute interval check, it will be equal to your proposal?
[18:57] <remote-lex> ant tor for 1 trying to ammoys be going tho anther peer but the peer that downloads the web page could get cort for downloading it its an p2p proxy client (well when the other soft is used)
[18:57] <KenMan> I am completely unsure what you mean.
[18:57] <toad_> or are you saying we need to completely reimplement it?
[18:57] <toad_> KenMan: currently a route is available if the average of the last 5 sends and this send is <= the MRI AND the time between this send and last send <= the MRI
[18:57] <KenMan> if you take out the check, you are open to flagrant abuse, which is unlikely in the near term
[18:58] <toad_> if we remove the latter check, then we are just checking the average
[18:58] <Diablo-D3> remote-lex: oh shuddup already.
[18:58] <Diablo-D3> remote-lex: freenet and tor == ilegal.
[18:58] <Diablo-D3> er, illegal
[18:58] <KenMan> let me ponder... of course I have a fixed rigid impl already in mind. Must remove it temporarily
[18:58] <Diablo-D3> damn keys are sticking
[18:58] <toad_> which shouldn't be a problem - if they want to mess around with the MRI, that's their problem
[18:59] <studdlybiggi> got a quick stupid question. I increased storeSize to 1024M thinking I am giving a bigger chunk of my harddrive for freenet, is that right? Or did I screw it up?
[18:59] <toad_> KenMan: well...
[18:59] <toad_> the first hit is at t+1
[18:59] <toad_> then the second hit is at 2t+2
[18:59] <verl> studdlybiggi: you have done a good thing :)
[18:59] <toad_> the average now allows the third hit to be at 3t-1
[19:00] <toad_> studdlybiggi: good
[19:00] <KenMan> yes, that is good. I think I get where yo uare going. Yes.
[19:00] <toad_> studdlybiggi: especially if you download big files, but in any case, big store == good
[19:00] <KenMan> It is a good approximation, depending on the actual MRI in use.
[19:00] <toad_> if the third hit is in fact at 3t+3, then the 4th hit can be as early as 4t-2
[19:00] <KenMan> if they are too small, it falls apart
[19:00] <studdlybiggi> ok then, wasnt sure if that was file sizes or node as a whole. I can spare even more but will it slow me down? I have about 100gig free, could give at least 25 but I dont want to slow down myself?
[19:01] <toad_> KenMan: of course the downside:
[19:01] <toad_> if we have no requests for a while, we can then have a whole bunch of them
[19:01] <toad_> depending on how long the average is
[19:01] <toad_> i think that's probably why i added that code
[19:01] <KenMan> that is the quasi-random nature you can experience with several requestors anyway
[19:01] <toad_> the question is, is that a problem?
[19:02] <toad_> yup
[19:02] <KenMan> it is not for me, it may be for you depending on how they compare to the periods you use for averaging things
[19:02] <toad_> okay
[19:02] <toad_> right now, PeerHandler uses the higher of the average and the individual interval to check
[19:03] <KenMan> if you are NOT changing MRI intervals , in that academic environment, you would still be okay
[19:03] <toad_> we'd have to change that to just use the average, and make sure that they're the same length
[19:03] <toad_> but it's quite feasible
[19:03] <toad_> hmmm
[19:03] <KenMan> i like it a lot... because it allows more flexibility in individual intervals
[19:04] <toad_> it won't produce 50% backoff without propagation... but WITH propagation, it probably ought to...
[19:04] <KenMan> while maintaining an average. But you must measure how well it conforms.
[19:04] <toad_> actually no
[19:04] <toad_> hmmm
[19:04] <toad_> will it or won't it?
[19:05] <toad_> R<-----------------------------big MRI------------------>F......R<------------------------big MRI----------------> where F is where it first becomes available
[19:05] <toad_> now, if we are trying to get equal incoming requests to outgoing requests, that does not mean that they will come in at the right times
[19:05] <KenMan> here is my extremely simplified model. One route. If I say you can have 300qph, and I never change it, then my model provides 50% backoff of that single route, as measured over a stretch of time
[19:05] <toad_> the only way we can make them come in at the right times is queueing
[19:05] <toad_> KenMan: I'm skeptical
[19:06] <KenMan> you CAN'T make them come in at any "right" time
[19:06] <KenMan> I know you are skeptical :)
[19:06] <toad_> if you have queries coming in every 5 seconds, that might like to go to route A, and route A has a fixed MRI of 300 seconds, then it will not produce 50% backoff
[19:06] <toad_> it'll be more like 90%
[19:06] <toad_> on THAT ONE NODE
[19:06] <KenMan> stop there
[19:06] <toad_> the only way to do it would be to propagate the WORST node in the RT's times
[19:06] <toad_> then you'd have good backoff...
[19:06] * toad_ stops
[19:07] * toad_ brb, please continue to explain, i will read backlog in 2 minutes
[19:07] <KenMan> you don't have them coming in every 5 seconds. On average maybe you do, but in reality you don't. The interarrival will always be different.
[19:08] <KenMan> If I give you cells worth 5 seconds, you are free to apply them at any interarrival values, so long as no interarrival exceeds 5 ( no wait, 10 ) seconds.
[19:09] <KenMan> i think i like your proposal above, it sounds like a move in the right direction in any case.
[19:10] <toad_> hmmm
[19:10] * toad_ back
[19:10] <toad_> KenMan: if a route has an MRI of 300 seconds, and it's a popular route, then it's likely that most of the time we will take up the route at 10 seconds or so
[19:11] <toad_> the result of this is that we then have a long wait until it becomes available again
[19:11] <KenMan> the average occurence will tend to center in the middle of the cell, so backoff would be 50%
[19:11] <toad_> KenMan: what makes you think that?
[19:11] * Daedalus007 (Daedalus00@ACAFF8CA.ipt.aol.com) has joined #freenet
[19:11] * studdlybiggi (~noneya@12-215-171-38.client.mchsi.com) Quit ()
[19:11] <KenMan> you are talking on big time scales, scale it down to milliseconds range and reconsider
[19:11] <toad_> I'd be interested to see an argument that with propagation the average occurrence will be in the middle
[19:12] <toad_> KenMan: i'm not interested in millis, most of my routes on my real nodes have loooong MRIs
[19:12] <KenMan> I've already made that argument, without pretty pictures or the right words... :(
[19:12] <toad_> KenMan: okay, how does it go?
[19:12] <toad_> I mean you might say "it just feels like it would"
[19:12] <toad_> but how do you go beyond that?
[19:13] <toad_> well you can prove it if you have only one route I suppose
[19:13] <KenMan> you just said it. You agreed earlier that with fixed unchanging MRIs you would never have problems... except that the interarrivals would still arrive at random intervals.
[19:13] <toad_> we have requests coming in on average every 100 seconds, we have a single route which we can send a request out every 100 seconds...
[19:13] <toad_> (on average)
[19:13] <Diablo-D3> hrm.
[19:13] <Diablo-D3> I just thought of a silly idea.
[19:14] <toad_> so the first one comes in
[19:14] <leex> heh
[19:14] <toad_> then we have no slack because we have no history, so we wait 100 secs, then become available
[19:14] <toad_> a request comes in at 130 secs
[19:14] <toad_> so we're +30
[19:14] <toad_> so the next request can be -30
[19:14] <toad_> i.e. at 200 secs
[19:14] <toad_> the next req comes in at 210 secs
[19:15] <KenMan> AND STILL MAINTAIN 100 sec "average" interarrival
[19:15] <KenMan> :)
[19:15] <toad_> the average is now 240/2 = 120
[19:15] <toad_> we're +20
[19:15] <toad_> the third one comes in at 280
[19:15] <toad_> so we're dead on now
[19:16] <toad_> hmmm
[19:16] <toad_> I suppose it's not exactly the same... it's different for bootstrapping e.g.
[19:16] <toad_> but with the cellular model, you have some interesting quantization artifacts...
[19:16] <toad_> lets say the first req is at 95% of the interval
[19:16] <toad_> and then they appear every 95%
[19:16] * DiabloD3 (~diablo@69.20.9.201) has joined #freenet
[19:16] <toad_> soon you'll have two in an interval
[19:17] <KenMan> one in 20 gets caught up and dropped...
[19:17] <remote-lex> lo DiabloD3 agane
[19:17] <remote-lex> :)
[19:17] <toad_> I don't know that it's any better that way though...
[19:17] * remote-lex (~le@spc1-warr1-3-0-cust198.bagu.broadband.ntl.com) Quit ()
[19:17] * rnoh-htat (~rnoh-htat@69-162-196-251.bflony.adelphia.net) has joined #freenet
[19:17] <toad_> I think it's the same basic principle
[19:17] <KenMan> but running at (1/.95) of the rate is the senders "choice" . I don't know that it is any better either.
[19:18] <toad_> okay, more complex example:
[19:18] * rnoh-htat (~rnoh-htat@69-162-196-251.bflony.adelphia.net) Quit (Client Quit)
[19:18] <toad_> we have 2 routes
[19:18] <DiabloD3> hrm.
[19:18] <DiabloD3> I just thought of a silly idea.
[19:18] <DiabloD3> whats the smallest a packet can be?
[19:18] <toad_> one has an MRI of 30 seconds
[19:18] <toad_> the other has an MRI of 10 seconds
[19:18] <toad_> DiabloD3: pretty small. what sort of packet?
[19:18] <DiabloD3> toad_: tcp
[19:18] <toad_> DiabloD3: 1 byte, but there's 66 bytes overhead
[19:19] <DiabloD3> I was thinking about what impact there would be if I send everything at, say, 1 byte at a time
[19:19] <toad_> KenMan: that means we can accept 8 queries per minute, which we send on to our downstream node
[19:19] <toad_> our requestor node
[19:19] <toad_> DiabloD3: you'd have a TON of overhead
[19:19] <toad_> 6700% !
[19:19] <toad_> well 6600%
[19:19] <toad_> KenMan: okay
[19:19] <DiabloD3> heh
[19:19] <toad_> I'm going to assume your model
[19:19] <DiabloD3> but what if I encrypted each byte?
[19:20] <toad_> you start both nodes off
[19:20] <toad_> both are in their first interval
[19:20] <toad_> a request arrives at t = 4 seconds
[19:20] <toad_> it goes to the node with the 30 second MRI, because that node is better
[19:20] <toad_> we now cannot send any more requests to that node until t = 30
[19:21] <toad_> a request arrives at t = 6 seconds
[19:21] <toad_> we send it to the 10 second MRI
[19:21] <toad_> a request arrives at 13 seconds
[19:21] <toad_> we send it to the 10 second MRI
[19:21] <toad_> a request arrives at 33 seconds
[19:21] <toad_> we send it to the 30 second MRI
[19:22] <toad_> note that we've just wasted the third interval for the 10 second node...
[19:22] <DiabloD3> heh
[19:22] <DiabloD3> that is kinda stupid
[19:22] <KenMan> no we didn't...
[19:22] <toad_> KenMan: with your model, we did
[19:22] <DiabloD3> toad_: what is MPI?
[19:22] * Diablo-D3 (~diablo@dialup-4.156.102.183.Dial1.Boston1.Level3.net) Quit (Client Quit)
[19:22] <toad_> with my model, we didn't
[19:22] <KenMan> no queries arrived at the 20-30 second interval
[19:22] <toad_> because the average carries over
[19:22] <KenMan> well your example is somewhat contrived :) so long as you are processing enough queries that the chosen cell sizes overlap, you will get to heaven.
[19:23] <toad_> KenMan: well, we'll get 8 queries incoming
[19:23] <toad_> we'll have to RNF some of them
[19:23] <toad_> that's my point
[19:23] <toad_> once the 30 second one is occupied, we can only take 1 req per 10 second interval
[19:23] <toad_> so if we then get one at t = 39, we have to RNF that one AS WELL
[19:24] <toad_> KenMan: we could simulate this model...
[19:24] <KenMan> it works out as the engine pumps harder
[19:24] <DiabloD3> GUYS
[19:24] <KenMan> yes, that is a good thought.
[19:24] <DiabloD3> what is an mpi?
[19:24] <toad_> KenMan: IMHO, we need queueing at some point to avoid RNFing ANY queries though
[19:25] <toad_> DiabloD3: big integer
[19:25] <toad_> integers bigger than 2^63
[19:25] <toad_> often used in crypto
[19:25] <KenMan> i won't slow you down from queueing by arguing with you about it :)
[19:25] <toad_> we use integers around 2^160 in freenet's crypto
[19:25] <DiabloD3> is 10 or 30 mpi better?
[19:25] <toad_> DiabloD3: ah
[19:25] <toad_> M*R*I
[19:25] <KenMan> It offers whole new queueing models, but I'm not quite comfortable with your existing model today.
[19:25] <toad_> is something else
[19:25] <toad_> MRI is minimum request interval
[19:26] <DiabloD3> ahh
[19:26] <DiabloD3> okay is 10 or 30 mri better?
[19:26] <KenMan> GIVE ME 10 , RIGHT NOW
[19:26] <KenMan> the funny thing with MRIs is that smaller means more, bigger means less
[19:26] <toad_> okay
[19:26] <DiabloD3> so, is 10 or 30 better?
[19:26] * toad_ codes the simulation
[19:26] * Daedalus007 (Daedalus00@ACAFF8CA.ipt.aol.com) Quit (Read error: 60 (Operation timed out))
[19:27] <toad_> all I have to do is select a bunch of intervals within the hypothetical minute
[19:27] <toad_> sort them
[19:27] <DiabloD3> is 30 better?
[19:27] <KenMan> use representative values like what we see today, please :)
[19:27] <toad_> and then process them according to the simple rules I gave above
[19:27] <DiabloD3> toad_: wait wait wait
[19:27] <DiabloD3> stop talking for a second
[19:27] <toad_> KenMan: I'll extend it later, I want something easy first
[19:27] <DiabloD3> you're saying 30 is better, right?
[19:27] <KenMan> depends on whether you are giving someone else the 30 or receiving it yourself ...
[19:28] <DiabloD3> well what is this argument about? getting or sending?
[19:28] <toad_> DiabloD3: in the hypothetical situation above, we have two nodes. The first one is really really good, but only accepts one request every 30 seconds. The second is less good and accepts one query every 10 seconds.
[19:28] <KenMan> making both sending and getting work a little better.
[19:28] <DiabloD3> toad_: then thats easy
[19:28] <DiabloD3> make a list
[19:28] <DiabloD3> from good to bad
[19:28] <DiabloD3> then when you do .. whatever you're doing, walk the list
[19:28] <DiabloD3> until you find a node that is free
[19:28] <DiabloD3> then use that node
[19:28] <KenMan> that's what we do today, except it is a long long walk :(
[19:29] <DiabloD3> ouch.
[19:29] <DiabloD3> well it seems to be the best way
[19:29] <KenMan> NGR evaluates 50 nodes, then discovers that only one or two are available for use
[19:29] <KenMan> :p
[19:30] <jokern> whoo, I hit the top :=)
[19:30] <jokern> Connections open (Inbound/Outbound/Limit) 200 (145/55/200)
[19:30] <jokern> Transfers active (Transmitting/Receiving) 161 (63/98)
[19:30] <DiabloD3> well what are mri usually like?
[19:30] <KenMan> i suppose it is worse for some people, like NGR computes values for 100 nodes, then discovers that only 3 are open...
[19:30] <DiabloD3> maybe you should group them by group
[19:31] <toad_> okay, i've created my array of event times...
[19:31] <DiabloD3> like, good, eh, not so good, bad, really bad, total shit
[19:31] <DiabloD3> then when you use a good, you must use an eh next
[19:32] <DiabloD3> then after eh, check if any good are available, then any eh are available, and if neither, use a not so good
[19:32] <DiabloD3> then check if good, eh, and not so good are available.
[19:32] * toad_ runs the sim over 300 seconds...
[19:32] <toad_> other values unchanged...
[19:32] <DiabloD3> so is my idea bad?
[19:37] <DiabloD3> yes? no?
[19:39] <KenMan> array of event times is a normal distribution with an average arrival time ? crank it up to 100s or more, the shorter the sim length, the more yours is at a disadvantage
[19:39] <KenMan> i mean 1000 seconds or more
[19:40] <KenMan> mine will show it's advantage as number of peers grows... this is good educational stuff ;)
[19:41] <KenMan> what we cannot control is the set of MRIs for our routes. We can control things like average and std dev of the MRIs we hand out - which affect the input events...
[19:41] * toad_ written sim, restarting...
[19:42] <KenMan> ;) i hope this proves useful. I may do my own , but then we might reach conflicting results ;)
[19:42] <KenMan> it is much easier to watch you for now :p
[19:42] <toad_> :)
[19:42] * toad_ suspects we will see significant numbers of RNFs even with the averaging
[19:43] <toad_> doh, didn't enable runMiscTests... 1sec...
[19:43] <KenMan> that is because, on a local level, RNFs that are allowed to retry will steal away open routes from other queries in any model.
[19:44] <KenMan> or do you speak in this academic / sim world ? looking solely at exhaustion ?
[19:44] <toad_> hmm, now to debug it...
[19:44] <toad_> KenMan: the sim is a best case scenario
[19:45] <KenMan> cool
[19:45] <KenMan> where routes don't change their MRIs ? ;)
[19:46] <toad_> KenMan: yes
[19:46] <toad_> it's simplified - 2 routes only
[19:46] <toad_> and they don't change their MRIs
[19:46] <toad_> the better one has the longer MRI
[19:46] <KenMan> okay for now... better meaning first choice for all queries.
[19:47] <toad_> okay, fixed the bug...
[19:47] <toad_> KenMan: it routes to the 30-second-mri node if it can
[19:47] <toad_> and then to the 10-second-mri node if it can
[19:47] <toad_> and then RNF
[19:47] <KenMan> gotcha
[19:48] <toad_> seems to average at 10 hits on node 1, 20 hits on node 2, 10 RNFs
[19:48] <DiabloD3> bleh
[19:48] <DiabloD3> fine, say my idea is bad
[19:48] <toad_> examples:
[19:48] <toad_> Totals: node 1 hit
[19:48] <toad_> 9 times, node 2 hit 22 times, and 9 RNFs
[19:48] <toad_> Totals: node 1 hit
[19:48] <toad_> 9 times, node 2 hit 17 times, and 14 RNFs
[19:49] <toad_> Totals: node 1 hit
[19:49] <toad_> 9 times, node 2 hit 22 times, and 9 RNFs
[19:49] <toad_> Totals: node 1 hit
[19:49] <toad_> 8 times, node 2 hit 19 times, and 13 RNFs
[19:49] <KenMan> the basic cell-frame versus cell+ math should guarantee the outcome
[19:49] * toad_ suspects there's some basic math reason for this ratio :)
[19:49] <toad_> KenMan: huh?
[19:49] <KenMan> just keep going...
[19:49] <toad_> I can simulate your way as well if you like
[19:50] <KenMan> did you implement cell+ without averaging ? or with averaging ?
[19:50] <toad_> with averaging
[19:50] <toad_> it'd be much worse without
[19:50] <KenMan> ah, okay. Yeah, you have to try the other way to be fair...
[19:50] <DiabloD3> INFO: Native CPUID library 'freenet/support/CPUInformation/libjcpuid-x86-linux.so' loaded from resource
[19:50] <DiabloD3> INFO: Optimized native BigInteger library 'net/i2p/util/libjbigi-linux-athlon.so' loaded from resource
[19:50] <DiabloD3> Caught java.lang.OutOfMemoryError running or seeding node
[19:50] <DiabloD3> java.lang.OutOfMemoryError
[19:50] <DiabloD3> Caught, in Main:
[19:50] <DiabloD3> java.lang.OutOfMemoryError
[19:50] <KenMan> meaning, my other way, not (without averaging) way
[19:50] <DiabloD3> java.lang.OutOfMemoryError
[19:50] <DiabloD3> well that cant be good
[19:51] <KenMan> how else could you make a comparison ?
[19:51] <toad_> KenMan: with averaging, without the single-interval-check
[19:51] <toad_> it'd be interesting to include the single-interval check, that might make it really bad :)
[19:52] <toad_> SEC Aug 10, 2004 1:10:08 AM (freenet.node.Main, main, NORMAL): Averages: Node 1:
[19:52] <toad_> 8.912, Node 2: 20.691, RNFs: 10.397
[19:52] <toad_> that's the average of 1000
[19:52] <KenMan> okay, i expect about the same results. Yours can average over a greater period than my "cell size" , i guess.
[19:52] <toad_> using crypto-safe random numbers
[19:53] <toad_> KenMan: which do you want me to check?
[19:54] <KenMan> whatever YOU are most interested in. That usually produces the best learning.
[19:54] * toad_ tries it with the current system - both the average and the single interval
[19:54] <toad_> this will probably really suck...
[19:55] <toad_> SEC Aug 10, 2004 1:13:24 AM (freenet.node.Main, main, NORMAL): Averages: Node 1:
[19:55] <toad_> 7.636, Node 2: 14.835, RNFs: 17.529
[19:55] <toad_> SEC Aug 10, 2004 1:13:58 AM (freenet.node.Main, main, NORMAL): Averages: Node 1:
[19:55] <toad_> 7.66, Node 2: 14.847, RNFs: 17.493
[19:55] <toad_> so it almost doubles the number of RNFs
[19:55] <toad_> given that particular incoming load
[19:56] <toad_> probably the incoming requests distrib is unrealistic? i dunno...
[19:56] <toad_> perhaps not
[19:56] <toad_> it probably IS the "random tiny chance at any instant"...
[19:59] <toad_> okay
[19:59] <toad_> this time, I'm using the model I used the first time, and trying to see if a kludge factor on the incoming requests makes much difference
[19:59] <KenMan> even though rate limiting is controlling things at a high level, it might be neat to measure average and std dev for interarrival times on a stable node with 50+ peers...
[19:59] <DiabloD3> okay
[19:59] <DiabloD3> I must be on crack
[19:59] <DiabloD3> it _is_ 8888 isnt it?
[20:00] <toad_> DiabloD3: what? the HTTP port? yes
[20:00] <toad_> should be
[20:00] <DiabloD3> wtf!?
[20:00] <DiabloD3> how do I make java use less memory?
[20:00] <KenMan> for browsing, yes. For inter-node communications it is that random number port (listenPort) in freenet.ini / freenet.conf
[20:00] <KenMan> DiabloD3: I've tried yelling at it, but that didn't seem to work... :(
[20:01] <DiabloD3> KenMan: no, it doesnt =/
[20:01] <leex> heh
[20:01] <leex> fs i still here
[20:01] <leex> and sleep and i ant at home
[20:02] <leex> sleep driveing
[20:02] <leex> sya tomoro maybe :)
[20:02] <KenMan> night ...
[20:02] <leex> intresting chats
[20:03] * leex is now known as leex-zzzz
[20:03] <KenMan> Current estimated load for rate limiting 56,723.9% , i guess I'll shut down frost here now...
[20:04] <toad_> SEC Aug 10, 2004 1:23:05 AM (freenet.node.Main, main, NORMAL): Averages for fact
[20:04] <toad_> or 1.0 - 40 in 300 seconds: Node 1: 8.921, Node 2: 20.687, RNFs: 10.392 = 0.2598
[20:05] <KenMan> it made my point, rather painfully. I think it is gonna take a long long time to come down off a 56000% limit :o
[20:05] <toad_> the good news is reducing the kludge factor does reduce the fraction of requests that RNF
[20:05] <toad_> the bad news is not fast enough
[20:05] <KenMan> what does the klidge factor do ?
[20:05] <DiabloD3> oh well
[20:05] <DiabloD3> I guess that ends the great freenet experiment
[20:06] <KenMan> hardly... there is much to learn here.
[20:06] <toad_> KenMan: hmm?
[20:06] <DiabloD3> I ment mine, KenMan
[20:06] <KenMan> stick with the sim work, vary the inputs and the outputs, and see what you learn...
[20:06] <toad_> SEC Aug 10, 2004 1:24:55 AM (freenet.node.Main, main, NORMAL): Averages for fact
[20:06] <toad_> or 1.0 - 320 in 300 seconds: Node 1: 75.299, Node 2: 165.113, RNFs: 79.588 = 0.2
[20:06] <toad_> 4871249999999998
[20:07] <toad_> we can reduce it to 5% by using a kludge factor of 50%
[20:07] <DiabloD3> the problem is java is a fucking peice of shit
[20:07] <DiabloD3> and the freenet authors are of questionable intelligence for using it
[20:07] <toad_> unfortunately, using a kludge factor of 50% means we will probably get a meltdown
[20:07] <KenMan> what is this "kludge factor" ?? something to do with generating the query times i guess ?
[20:07] <toad_> DiabloD3: :)
[20:07] <toad_> KenMan: well, in theory, we can accept N requests in the target period
[20:08] <toad_> based on the MRIs of the routes
[20:08] <toad_> we multiply that by the kludge factor
[20:08] <DiabloD3> well, someone tell me when freenet is rewritten in C, okay?
[20:08] <toad_> to reduce the number of requests and improve the backoff situation
[20:08] <toad_> DiabloD3: never happen. Well not for 10 years.
[20:08] <toad_> DiabloD3: what's your specific java problem?
[20:08] <DiabloD3> then freenet will suck for 10 years.
[20:08] <toad_> DiabloD3: we may have a native code version via GCJ at some point though
[20:08] <KenMan> DiabloD3: it was done , the 'Entropy' Project. Go google it. btw it just died recently...
[20:08] <DiabloD3> it says java.lang.OutOfMemoryError a few times then refuses to work.
[20:08] <toad_> Entropy is not Freenet
[20:08] <toad_> Entropy sucks for other reasons
[20:09] <toad_> DiabloD3: windows or linux?
[20:09] <DiabloD3> linux
[20:09] <toad_> DiabloD3: I apologize, I was supposed to be fixing that...
[20:09] <DiabloD3> gjc would be a huge improvement here
[20:09] <toad_> DiabloD3: okay, edit the start-freenet.sh
[20:09] <toad_> where it says -Xmx128M, change it to -Xmx192M
[20:09] <toad_> or something like that
[20:09] <toad_> and try again
[20:09] <DiabloD3> toad_: that makes it even worse
[20:09] <toad_> DiabloD3: it does?
[20:09] <DiabloD3> Yeah
[20:09] <toad_> it won't OOM
[20:10] <DiabloD3> I was playing with that earlier
[20:10] <toad_> okay, then just cut your seednodes.ref in half
[20:10] <DiabloD3> no, it just makes my box almost die
[20:10] <DiabloD3> the kernel even oom killed it once
[20:10] <toad_> DiabloD3: what happens if you give it more RAM as above?
[20:10] <toad_> ahh, how much RAM and swap do you have?
[20:10] <DiabloD3> 256 and 256.
[20:10] <KenMan> and how much RAMchips does you have ?
[20:10] <toad_> DiabloD3: hmmm
[20:10] <toad_> that's odd
[20:11] <KenMan> hmmm... on windows ? linux ?
[20:11] <toad_> what JVM are you using? you said 1.5 uses tons of RAM?
[20:11] <toad_> tried 1.4.2 latest?
[20:11] <DiabloD3> KenMan: 8 on one so-dimm.
[20:11] <DiabloD3> 1.4.2_05 now
[20:11] <KenMan> hehe
[20:11] <DiabloD3> 1.5 was even worse
[20:12] <toad_> KenMan: bottom line: IMHO a kludge factor will probably lead to freezedown. Without one we're looking at 25% of incoming requests RNFing. Granted, it's an unrealistic simulation, but...
[20:12] <toad_> 35% with the current code
[20:12] <KenMan> it has some meaning, no doubt. Keep at it in your spare time, you will learn a lot. I will try to build something reasonable on my end...
[20:12] <toad_> 25% with taking the thingy out
[20:12] <DiabloD3> you know what the sad part is?
[20:13] <KenMan> ...see what i can learn also
[20:13] <DiabloD3> the sad part is, if it was written in c, it would take like 8 megs of memory
[20:13] <KenMan> DiabloD3: shhh !!!
[20:13] <toad_> DiabloD3: the major concern with freenet is NOT its resource usage at the moment
[20:13] <toad_> it's making the bloody thing work at all
[20:13] <KenMan> heh
[20:14] <DiabloD3> toad_: feh
[20:14] <toad_> KenMan: okay, what's the next step?
[20:14] <DiabloD3> well you'd get more devs if it wasnt written in a stupid language
[20:14] <toad_> I'm not sure how to test the propagation code...
[20:14] <toad_> DiabloD3: perhaps but it'd take a year to port it
[20:14] <KenMan> i dunno... i was thinking about stable... is there a limit to the number of retries for a single query ??
[20:14] <toad_> DiabloD3: also, Java is the COBOL of the 90s
[20:14] <DiabloD3> a year?
[20:14] <DiabloD3> if I knew java, I could port it in like 5 minutes.
[20:14] <toad_> half of all devs who have worked commercially know it
[20:15] <toad_> also, we'd lose some of the existing devs
[20:15] <toad_> DiabloD3: it's 165kloc
[20:15] <KenMan> but it is worthy of maybe 89kloc ...
[20:15] <toad_> KenMan: hmm?
[20:15] <KenMan> ;)
[20:15] * lolo-laptop (~lostlogic@68.251.84.226) Quit ("Client exiting")
[20:15] <toad_> re retries - hmm?
[20:15] <DiabloD3> toad_: I write that much an hour!
[20:15] <KenMan> well, ya know, the off "this module is deprecated" and what not...
[20:15] <toad_> DiabloD3: and it'd then take you the next year to debug it ;)
[20:16] <DiabloD3> okay, fine
[20:16] <toad_> in any case I'd want to use C++
[20:16] <DiabloD3> Ill write it in objc instead
[20:16] <KenMan> oh, is there a fixed limit (in 5090) on how many retries a single request can attempt ?
[20:16] <toad_> but we might well lose some devs over it
[20:16] <DiabloD3> not c++!
[20:16] <DiabloD3> objc > c++
[20:16] <DiabloD3> objc > c++
[20:16] <DiabloD3> objc > c++
[20:16] <toad_> and we'd spend a year arguing over whether to link in Boehm GC ;)
[20:16] <toad_> DiabloD3: yeah, 'cos everyone knows objc!
[20:17] <DiabloD3> toad_: objc > you
[20:17] <toad_> DiabloD3: lets rewrite it in m68k assembler and be done with it
[20:17] <toad_> KenMan: no
[20:17] <DiabloD3> sad part is I could.
[20:17] <KenMan> Could you institute one easily ? Like a value of 1 ? Allow each request no more than a single retry ??
[20:17] <toad_> KenMan: should i simulate anything else?
[20:17] <toad_> yes, if I wanted to, but I don't
[20:17] <KenMan> if you want to, or leave it for later...
[20:17] <KenMan> re:sim
[20:18] <toad_> because I don't think it would make any difference
[20:18] <toad_> as i have explained
[20:18] <KenMan> I really really think it would. In place of coding changes , could we solicit some numbers for max sFC from people ?
[20:18] <toad_> what we need to do is find out whether the propagation code would make a difference
[20:18] <toad_> KenMan: what's the point?
[20:19] * DiabloD3 (~diablo@69.20.9.201) has left #freenet
[20:19] <toad_> if they say 12, I wouldn't be convinced
[20:19] <toad_> (they probably can't, given HTL)
[20:19] <KenMan> then i can't get any further with you at this point in time ;)
[20:19] <toad_> KenMan: you have not given me anything concrete w.r.t. making RNFs fatal
[20:19] <toad_> KenMan: I'm interested in propagation and queueing though
[20:20] <KenMan> most RNFs probably only go "one deep" and immediately turn around after hitting the brick wall
[20:20] <toad_> how do i simulate queueing?
[20:20] <KenMan> oh hell...
[20:20] <KenMan> i won't offer advice on that one !!
[20:20] <toad_> heh
[20:20] <KenMan> although it shouldn't be hard ... :)
[20:20] <toad_> it'd be a minor pita though
[20:20] <KenMan> in the real code, yes. I should think so.
[20:21] <toad_> but i think as far as Deep Magic goes, we need to find out whether propagation as presently implemented is a problem
[20:21] <KenMan> I just want to make sure we try all our other options first. I mean, these MRI schemes are strikingly similar to queueing models...
[20:21] <toad_> that means greatly expanding the new-unstable-testnet
[20:21] <KenMan> good enough. I'm with you on that.
[20:22] <toad_> okay
[20:22] <KenMan> What about the local queries effect though ? I'm still at 9293% load here, after stopping frost some time ago.
[20:22] <toad_> I'll commit, turn off snapshots, upload the jar and seednodes, and ask for users
[20:22] <toad_> KenMan: well that's mostly because there is virtually no traffic on our testnet imho
[20:22] <KenMan> good enough, let's go forward.
[20:23] <KenMan> (until we realize it has all been sideways, anyway) ;)
[20:23] <toad_> I might be wrong about that... but I doubt frost would sustain 80qpm over HOURS
[20:23] <toad_> KenMan: :)
[20:23] <toad_> tomorrow I need to fix the seednodes memory usage
[20:23] <toad_> somehow
[20:23] <toad_> before doing any more deep stuff
[20:24] <KenMan> only read in the first X items, or a random X number from which to further choose...
[20:24] <KenMan> serial disk reads, set a limit on how many you take in
[20:25] <toad_> KenMan: no
[20:25] <toad_> we have to choose N random, valid items
[20:25] <KenMan> funny how quick I am to come up with suggestions off-the-cuff
[20:25] <toad_> that's the problem
[20:25] <toad_> that means either we count them and validate them in the first pass and then read them
[20:25] <toad_> or we store them in ram in a more compact format and use 1 pass
[20:26] <toad_> or we read them and store offsets
[20:26] <KenMan> well, I have confidence you will solve it in any case :)
[20:26] <KenMan> it is a good sign that we gather so many seeds, no ?
[20:28] <KenMan> I still haven't convinced you that RNFs which result in more outgoing queries for a single request hurt the backoff makeup. We need a term for the table of collective backoff/avaiable values.
[20:28] <toad_> ugh, eclipse wants to add lib/freenet*.jar
[20:29] <KenMan> anyway, chat with you later... i must go forage now (grocery shop).
[20:55] * toad_ looking forward to it
[20:56] * toad_ bbl zzz too
[20:56] <toad_> you'll be around late tomorrow? in the morning i need to fix the seednodes somehow
[20:56] <toad_> probably i'll do a 2-pass read... create an array of items with indexes in them
[20:57] <toad_> then do random seeked reads
[20:57] <toad_> have to copy it beforehand to avoid it being modified
[20:57] <toad_> also it doesn't really belong in the RT
[20:58] <toad_> oh and I'm 90% sure i'm going to aikido tomorrow
[20:59] <toad_> i might then go on wednesday or friday as well
[20:59] <toad_> on thursday i have another engagement which occupies about the same time
[20:59] <toad_> bbl
[21:14] * Zorix (Brandon@fl-64-45-226-120.dyn.sprint-hsd.net) Quit (Read error: 110 (Connection timed out))
[21:30] * lostlogic (~lostlogic@node-402418b2.mdw.onnet.us.uu.net) has joined #freenet
[21:51] * lostlogic (~lostlogic@node-402418b2.mdw.onnet.us.uu.net) Quit (Read error: 110 (Connection timed out))
[22:09] * rnoh-htat (~rnoh-htat@69-162-196-251.bflony.adelphia.net) has joined #freenet
[22:10] * rnoh-htat (~rnoh-htat@69-162-196-251.bflony.adelphia.net) Quit (Remote closed the connection)
[22:18] * supermann (~kdfjk@the.dogs.bollo.cx) has joined #freenet
[22:29] * Zorix (Brandon@fl-64-45-226-120.dyn.sprint-hsd.net) has joined #freenet
[22:42] * supermann (~kdfjk@the.dogs.bollo.cx) has left #freenet
[22:51] * bluephile (bluephile@69-160-205-193.clvdoh.adelphia.net) Quit (Client Quit)
[22:52] * salahx (salahx@sc1-24.217.174.147.charter-stl.com) has joined #freenet
[22:59] * thelema lost internet earlier
[22:59] <salahx> well welcome back then :)
[23:02] * rnoh-htat (~rnoh-htat@69-162-196-251.bflony.adelphia.net) has joined #freenet
[23:02] * rnoh-htat (~rnoh-htat@69-162-196-251.bflony.adelphia.net) Quit (Client Quit)
[23:17] <salahx> stable is really working well
[23:19] <jabawok_h2> yep apart from high RNFs.. managable thouhg
[23:20] <salahx> not getting many RRNF
[23:20] <salahx> At least nowhere near as many as before
[23:20] <salahx> its "tolerable"
[23:20] <jabawok_h2> yep
[23:20] <jabawok_h2> hopfully the work toad is doing with unstable-experimental will show some promise..
[23:21] <salahx> yeah
[23:21] <salahx> I'm gonna join it
[23:21] <salahx> I just need to download a few more things off "stable"
[23:21] <salahx> I have some incompelte splitfiles while I was on "unstable", adn am trying to get them fom the "other side" of the network...
[23:28] <thelema> does anyone know what toad is trying to do with this incoming - coutgoing mumbo jumbo?
[23:28] <salahx> Its something KenMan suggested
[23:28] <thelema> What's the big deal if there's more outgoing than incoming requests?
[23:29] <thelema> I guess there's sora pressure to specialize...
[23:29] * thelema tries reading IRC log
[23:29] <salahx> The idea is kinda pointless to accept packets than we can actualy route
[23:36] <thelema> agreed.
[23:36] <thelema> I guess the assumption is that our average outgoing is our max outgoing
[23:38] <salahx> of course we react to mRI increase by increaing out mRI, then oru neighbot node will increase their mRI, which in turn...
[23:39] <salahx> Hence the "network freezedown" warning :)
[23:40] <thelema> yes... we've done that before
[23:45] <salahx> this will be interseting
[23:45] <salahx> Our misson, if we choose to accept it, is to run this unstabe build adn see what happens
[23:46] <salahx> In teh event the netowrk get totally hosed, toaad will disavow any knowlege of the build
[23:47] <salahx> This message wqill slef-destruct in 3 seconds...
[23:49] <thelema> maybe "the network will self-destruct in 3 seconds"
[23:49] <salahx> maybe :)
[23:49] * Ash-Fox (Hal-9000@abd81.neoplus.adsl.tpnet.pl) has joined #FreeNET
[23:58] <Hamled|Erp|> anyone else getting nothing but RNF? (only pages I can see are the local one and the GPL, everythign else doesn't even connect to another node) Using 5090
[23:58] <Hamled|Erp|> basically started when I switched to unstable, but I went back to stable and it still isn't working
[23:58] <salahx> did you reseed ?
[23:58] <jabawok_h2> what does the connections page say?
[23:59] <Hamled|Erp|> no connections
[23:59] <jabawok_h2> what seednodes.ref did you use?
These logs were automatically created by Jay Oliveri with his gimp hapi on irc.freenode.net.
