Timestamps are in GMT/BST.
[0:00] * MrNaughty (MrNaughty@d198-166-55-147.abhsia.telus.net) has joined #freenet
[0:32] * Ash-Fox (Hal-9000@aaq112.neoplus.adsl.tpnet.pl) Quit (Client Quit)
[0:34] * Ash-Fox (Hal-9000@aaq112.neoplus.adsl.tpnet.pl) has joined #FreeNET
[1:42] * jmercier (~jmercier@4.10.164.111) has joined #freenet
[1:43] * protok0l (~hfdh@cpe-24-24-240-215.socal.rr.com) has joined #freenet
[2:10] * Ash-Fox (Hal-9000@aaq112.neoplus.adsl.tpnet.pl) Quit (Nick collision from services.)
[2:10] * Ash-Foxeus (Hal-9000@aaq112.neoplus.adsl.tpnet.pl) has joined #FreeNET
[2:31] * Ash-Fox (Hal-9000@aaq112.neoplus.adsl.tpnet.pl) has joined #FreeNET
[2:31] * Ash-Foxeus (Hal-9000@aaq112.neoplus.adsl.tpnet.pl) Quit (Client Quit)
[2:39] * Ash-Fox (Hal-9000@aaq112.neoplus.adsl.tpnet.pl) Quit (Client Quit)
[2:39] * Ash-Foxeus (Hal-9000@aaq112.neoplus.adsl.tpnet.pl) has joined #FreeNET
[3:28] * TheSeeker (~Fridlekh@cpe-65-28-158-211.socal.rr.com) has joined #Freenet
[4:12] * protok0l (~hfdh@cpe-24-24-240-215.socal.rr.com) Quit ("Let's make like a stoner and roll this joint")
[4:21] * TheSeeker (~Fridlekh@cpe-65-28-158-211.socal.rr.com) Quit (Read error: 60 (Operation timed out))
[4:31] * tyesinclair (~tyesincla@170.252.64.1) has joined #freenet
[4:31] * tyesinclair (~tyesincla@170.252.64.1) has left #freenet
[4:40] <iip_i2p> <Newsbyte> toad (or whomever), is there some info I can send you about my (crappy) nodeconnection that's of interest?
[5:16] <KenMan> outbound bandwidth limit and number of peers for starters...
[5:23] <KenMan> even though everything is erratic as hell, I *am* resolving a higher volume than normal. But 'successRatio' is no better. I don't know what to make of 5086 so far :(
[5:27] <KenMan> most routes have higher min & max xfer rates than I'm used to seeing. I guess that is a good thing...
[5:29] * TheSeeker (~Fridlekh@cpe-65-28-158-211.socal.rr.com) has joined #Freenet
[5:30] <sanity> hi kenman
[5:30] <KenMan> hi sanity
[5:31] <KenMan> that's right, you need graphs. Ones that show good things. I may have to go back in time then :)
[5:31] <sanity> KenMan: did you get that email i sent yesterday (or perhaps the day before)?
[5:31] <sanity> well, they don't have to show good things, only interesting things
[5:31] <sanity> but a few good things would be nice :-)
[5:31] <KenMan> what you want to demonstrate ? smooth query rates ? smooth bw rates ?
[5:32] <sanity> either/or
[5:35] <KenMan> how about the testnet barcode graphic - you could say almost anything about this one - http://mywebpages.comcast.net/jkcorson/FreeNet/60167barcodes.png
[5:35] <KenMan> one where I was attempting a better visual for specialization - http://mywebpages.comcast.net/jkcorson/ringspace.png
[5:36] <KenMan> another one, before I decided barcodes are best - http://mywebpages.comcast.net/jkcorson/barspace.png
[5:36] * guido^pe (~unknown@dsl-213-023-243-038.arcor-ip.net) has joined #freenet
[5:38] <sanity> KenMan: wow, ringspace is cool - what is it? :-)
[5:39] <KenMan> it was supposed to show the specialization of each node in a testnet . as keyspace is 'continuous' i figured a ring might beat a linear barcode...
[5:40] <KenMan> i got distracted by making this stuff look like molecular research or somesuch nonsense, and reverted to good old barcodes :)
[5:40] <sanity> it was a good idea
[5:40] <sanity> the testnet barcode graphic - wasn't that forced specialization?
[5:41] <sanity> what does it really say thats useful?
[5:41] <KenMan> it started out that way, and grew to show those requests that had the final HTL extended...
[5:42] <KenMan> it says that requests were indeed being clustered around that pre-configured sector , although it is a DS representation, not query rep
[5:44] <sanity> the ring thing shows a pretty even distribution of data - no?
[5:48] * Hory (~Miranda@81.196.25.110) has joined #FreeNet
[5:48] <KenMan> well, yeah, I forget what I was doing there. I think I was just slinging keys randomly, and hoping for emergence, so no surprise... :p
[5:50] <KenMan> i was just trying out different visuals on a set of data stores . The barcode for those rings showed minor clustering, much like the original barcodes. In the beginning, you would already start to identify dark spots.
[5:51] <KenMan> Some dark spots would grow and some would fade.
[5:58] <sanity> you animiated it?
[5:58] <sanity> animated even
[5:59] <KenMan> sanity, I'm too lazy to cherry-pick :o you choose - http://mywebpages.comcast.net/jkcorson/FreeNet/index.html
[5:59] <sanity> ooh, can you run a thumbnail generator on it? or zip them up and send them to me?
[6:00] <KenMan> you lazy bastard! okay.
[6:00] <sanity> 145xfers.png - why is the xfers incoming line going up?
[6:00] <KenMan> I mail bomb you , infidel !!
[6:00] <KenMan> toad says it is because the reporting is screwed up (since 60145)
[6:01] <KenMan> we tolerate things like this because he tends to fix more important matters ;)
[6:02] <sanity> hmz, do we have anything that demonstrated performance before rate limiting? a before/after would be nice
[6:03] <KenMan> i don't. I started charting because I wanted to see if it worked ...
[6:03] <sanity> ok
[6:04] <oierw> http://mywebpages.comcast.net/jkcorson/FreeNet/60145.png scares me
[6:04] <KenMan> actually, I was trying to verify bandwidth limiting, but query rate limiting was already impl'ed
[6:04] <oierw> seems to be better in 60167
[6:05] <sanity> KenMan: why does that scare you?
[6:05] <oierw> err.. that was to me
[6:05] <KenMan> nothing scares me, that was oierw
[6:05] <oierw> i guess it's the beginning onf the incoming xfer line bug
[6:05] <oierw> n/m
[6:06] <KenMan> the initial spike is normal , the dropouts near the end are also normal (for my buggy computer)
[6:07] <oierw> wow. I think i might be getting tired. outgoing xfers.
[6:08] <sanity> oh, sorry
[6:08] <KenMan> must be around 2pm on oierw-time
[6:09] <KenMan> sanity - where shall I mail-bomb you with these images ??
[6:09] <sanity> ian at locut dot us
[6:09] <KenMan> (it is a zip file)
[6:13] <KenMan> 1M mailbomb sending ........ sent. (my ISP [bw]rate limits email sends !)
[6:17] <KenMan> 60116mri1.png - this one shows how MRI can flux ...
[6:20] <KenMan> hehe - stats50.png ... I was gettin' bizzaee
[6:29] <sanity> KenMan: got your email
[6:30] <KenMan> well, have fun with it ! don't hesitate to make up interpretations, I often have to do that for some of these graphs... :o
[6:32] <KenMan> sanity: you are way,way off LA time... you are gonna crash at DEFCON... "And here, you can see ZZZZZ"
[6:40] * sanity is immune to jetlag these days
[6:41] <oierw> KenMan: do you have an automated way of making these graphs?
[6:52] * unknown_ (~unknown@dsl-213-023-238-173.arcor-ip.net) has joined #freenet
[6:54] <KenMan> automated ? I guess so, it is all scripted up. I just say 'plot' or 'plot2' and , voila, I get a graph.
[6:55] <KenMan> ...although lately, I don't like what gets displayed.
[6:58] * ncrfgs (~ncrfgs@host15-93.pool8249.interbusiness.it) has joined #freenet
[6:59] <ncrfgs> scuse me
[6:59] <ncrfgs> if I bother
[7:00] <ncrfgs> could anyone pls tell me
[7:00] <ncrfgs> whether freenet has always been written
[7:00] <ncrfgs> in java?
[7:04] <KenMan> so far as I know, it has always been done in java, from the beginning.
[7:04] <KenMan> there is no other way to explain the huge splort of code that is FreeNet
[7:05] <ncrfgs> I don't mean to start a flame or to troll but I don't understand then
[7:05] <ncrfgs> Freenet is Free Software
[7:05] <ncrfgs> but if I want to run it
[7:05] <ncrfgs> I have to install proprietary software
[7:05] <ncrfgs> isn't it paradoxal?
[7:07] <oierw> well... sort of
[7:07] <ncrfgs> =(
[7:07] <oierw> freenet was initially written for blackdown
[7:07] <oierw> no features were allowed that were not blackdown compatible
[7:07] <oierw> however, blackdown is buggy
[7:08] <oierw> and the project was dead for a while
[7:08] <oierw> it picked back up, i'm not sure if it still is alive (checks)
[7:08] <ncrfgs> I always heard of freenet and I always really do appreciate its philosophy, today I wanted to download it and compile it...
[7:08] <ncrfgs> and I discovered it was written in java
[7:09] <KenMan> wow. Seagate to offer 5 year warrantees on all hard drives. Just a year or two back, all mfg's shrunk their warranty periods to 1 year, except on the premium models (with big caches).
[7:10] <KenMan> go seagate! I am a Seagate-only customer for 3 years now...
[7:10] <ncrfgs> tnx anyway, scuse me if I bothered you
[7:10] <oierw> well... it's really that java didn't work unless you were using a proprietary system
[7:10] * guido^pe (~unknown@dsl-213-023-243-038.arcor-ip.net) Quit (Read error: 110 (Connection timed out))
[7:11] <oierw> classpath started adding new classes that weren't java1.1 compatible
[7:11] <oierw> so freenet adopted them
[7:11] <oierw> i'm looking at the feasibility of freenet and blackdown right now
[7:11] <toad_> <ncrfgs> I always heard of freenet and I always really do appreciate its philosophy, today I wanted to download it and compile it...
[7:11] <toad_> <ncrfgs> and I discovered it was written in java
[7:11] <toad_> :<<
[7:12] * toad_ needs to get together with robilad and debug freenet-under-kaffe
[7:12] * robilad[afk] is now known as robilad
[7:12] <ncrfgs> toad_, ?
[7:12] <toad_> unfortunately i need to do 27,000 other urgent things on freenet too !
[7:12] <robilad> hi toad_ :)
[7:12] <oierw> oh yeah. kaffe too :p
[7:12] <oierw> blackdown actually seems alive and kicking. might be interesteing to see how far it gets before it dies
[7:12] <toad_> <oierw> i'm looking at the feasibility of freenet and blackdown right now
[7:12] <toad_> blackdown isn't Free Software
[7:13] * oierw has been searching for a license
[7:13] <toad_> it's slightly less horrible than Sun iirc, but only slightly
[7:13] <robilad> the big swing/awt merge is done, and I've actually mostly resynced with classpath.
[7:13] <toad_> it's closely based on the Sun code
[7:13] <toad_> we were using what works with Kaffe
[7:13] <toad_> not with Blackdown
[7:13] <toad_> as our basis
[7:13] <toad_> then we added NIO... :<
[7:13] <oierw> well, I'm all messed up. I went with the first name I rememebred. I'm sorry
[7:14] <robilad> well, nio's got some way to go in classpath. we *need* more nio hackers.
[7:14] <toad_> robilad: there should be some priority to getting freenet working with Kaffe. I've got to fix some annoying incredibly important connections bug, but after that, I'm available. I'll mail you when I have done that.
[7:14] <robilad> the basic stuff is there, but getting it all to work as expected is no small feat.
[7:15] <robilad> toad_: thanks. i'll get in touch with man-di so that we can have some debugging fun.
[7:15] <robilad> btw, tomcat5 now runs on kaffe
[7:15] <toad_> cool
[7:15] <toad_> what doesn't?
[7:15] <toad_> apart from freenet?
[7:15] <toad_> and does tomcat5 run on gcj?
[7:15] <robilad> with a one-liner change to the startup script to add a jar file to classpath as we don't have a real Manifest-classpath-aware class loader yet
[7:16] <robilad> gcj from cvs maybe.
[7:16] <robilad> haven't tried.
[7:16] <toad_> will freenet run on gcj very soon after it runs on kaffe?
[7:17] * toad_ thinks making Freenet run on Classpath is an important 0.6 criterion
[7:18] <toad_> actually I've got to deal with the why-on-earth-are-half-my-outgoing-MRIs-600k? bug too
[7:18] <toad_> oh well
[7:18] <toad_> robilad: contact me when you're free for a while
[7:19] <toad_> we can escalate to man-di when we find something concrete
[7:19] <robilad> toad_: hopefully, yes
[7:19] <robilad> i've made a few changes to kaffe meanwhile.,
[7:19] <robilad> the default is now the slow, working everywhere pure java big math, but gmp-math gets compiled in, if available.
[7:19] <toad_> ahhh
[7:20] <robilad> then you just run kaffe with kaffe -Xnative-big-math to run the things nicely fast ;)
[7:20] <toad_> hrrm
[7:20] <oierw> "Anecdotes: Should work, since Oskar Sandberg, one of the developers, develops using kaffe."
[7:20] <toad_> well I suppose it's better to have SlowerThanSunShitSlowMath than no math
[7:20] <robilad> I've also merged in dnsjava for fast async dns
[7:20] <toad_> ahhh
[7:20] <toad_> you have to use a special API for that though?
[7:20] <robilad> nope, just a property ;)
[7:20] <toad_> heh, same thing...
[7:21] <toad_> robilad:L
[7:21] <toad_> ummm above
[7:21] <robilad> kaffe -Dorg.kaffe.dns=java.net.DNSJavaInetAddressImpl -Xnative-big-math would give you both ;)
[7:21] <toad_> <robilad> then you just run kaffe with kaffe -Xnative-big-math to run the things nicely fast ;)
[7:21] <toad_> it should auto-detect it
[7:21] <robilad> hm .... true.
[7:21] <toad_> in fact, there's no reason not to just dynlink it in if it happens to be available on build
[7:22] <robilad> i'm still working out the details of making it useable ;)
[7:22] <toad_> with an option to staticlink it for distribution
[7:22] <toad_> you COULD load it at run tim
[7:22] <toad_> e
[7:22] <toad_> but I don't see why you shouldn't just link to it if it'
[7:22] <toad_> it's available
[7:23] <toad_> and fail if you don't have it, like 99% of other linux software
[7:23] <robilad> as i'd like to switch us to pure classpath with kaffe-speedy-implementations available as dropins on platforms that support it.
[7:23] <toad_> "what part of link dependancy don't you understand?" :)
[7:23] <robilad> heh ;)
[7:23] <toad_> well, I'm just saying, there is absolutely no reason not to depend on it by default if it was there at configure time
[7:23] * unknown_ (~unknown@dsl-213-023-238-173.arcor-ip.net) Quit (Remote closed the connection)
[7:24] <robilad> i need to play with bootclasspath, though, for the things to work, atm.
[7:24] <toad_> what you do NOT want is your math to be slower than Sun's without some unknown command line option
[7:24] <toad_> it should just drop in and work at maximum speed
[7:25] <robilad> yep. flame me on list, so that i don't forget to fix it ;)
[7:27] <robilad> but i'd be interested in the perfomance effect of the async dns stuff. too
[7:27] <robilad> does freenet use dns much?
[7:28] <Pascal> "making Freenet run on Classpath" - what does that mean?
[7:28] <robilad> GNU Classpath -> free runtimes -> everywhere ;)
[7:31] <toad_> robilad: ncrfgs would like to know how to use Kaffe in Mozilla to run applets :)
[7:32] <toad_> robilad: more than it should :|
[7:32] <robilad> toad_: not yet, afaik, though gcjwebplugin now works out of the box with kaffe from CVS
[7:32] <robilad> but i haven't tried the actual plugin yet
[7:33] <robilad> it's on my list of things to merge in for 1.1.5
[7:33] <robilad> the only major thing left out now is jacorb & gcjwebplugin ;)
[7:41] <toad_> bbiab
[8:14] <Iakin3> [02:36] * toad_ would prefer that a search client is integrated into fproxy
[8:14] <Iakin3> Same here
[8:15] <Iakin3> But 'on the side' of it of course
[8:18] <toad_> Java Applets: Write once, debug everywhere
[8:18] <toad_> Iakin3: what do you mean by on the side?
[8:20] * toad_ has been messing around with the Inland Revenue's Microsoft VM specific "view your tax return" option :(
[8:20] * toad_ predicts that it may just be fixed by the time of the submission deadline
[8:20] * greycat (~wooledg@192.35.79.70) has joined #freenet
[8:23] * TheSeeker (~Fridlekh@cpe-65-28-158-211.socal.rr.com) Quit (Read error: 60 (Operation timed out))
[8:23] <toad_> nice, my node is sending 600000.0ms MRIs to EVERY node for the last nearly an hour
[8:25] <toad_> globalQuo
[8:25] <toad_> ta: 179.7524606895061, totalRequests: 320.159418439388, totalRequestsThisNode: 5
[8:25] <toad_> .7455753026034, thisNodeMinQuota: 0.6245317926924481, thisNodeQuota: 0.624531792
[8:25] <toad_> 6924481, minRequestInterval: 600000.0
[8:25] <toad_> typical example...
[8:25] <wind789> stable?
[8:25] <toad_> unstable with some patches
[8:25] <wind789> aahh
[8:26] <toad_> hmmm
[8:26] <toad_> routingTime is 1499ms
[8:26] <toad_> therefore load is ~ 150%
[8:26] <toad_> but we have 50% idle CPU!
[8:27] <wind789> I see, is unpatched unstable currently broken?
[8:27] <toad_> I don't know
[8:27] <wind789> hehe
[8:27] <toad_> both stable and unstable have a fairly serious bug to do with connections failing
[8:27] <wind789> I see
[8:29] <toad_> hmm
[8:29] <toad_> estimated incoming requests per hour: 300
[8:29] <toad_> averaged rate limiting load: 1.7
[8:30] <toad_> => globalQuota = 300/1.7 ~= 170
[8:33] <wind789> bbl
[8:38] <toad_> bbl
[8:40] * ncrfgs (~ncrfgs@host15-93.pool8249.interbusiness.it) Quit ("Leaving")
[8:40] * ncrfgs (~ncrfgs@host15-93.pool8249.interbusiness.it) has joined #freenet
[8:40] <toad_> bbl, ~~ 1-2 hours
[8:41] * ncrfgs (~ncrfgs@host15-93.pool8249.interbusiness.it) Quit (Client Quit)
[8:44] * Hory (~Miranda@81.196.25.110) Quit ("Miranda IM! Smaller, Faster, Easier. http://miranda-im.org")
[8:58] <robilad> toad_: thanks for the mail!
[9:06] * lolo-laptop (~lostlogic@68.251.84.226) has joined #freenet
[9:12] <Iakin3> http://mywebpages.comcast.net/jkcorson/FreeNet/5086c.png
[9:12] <Iakin3> Whoa.. interesting
[9:16] * thelema|away is now known as thelema
[9:18] <thelema> Iakin: you're going way over 150 queries per minute?
[9:21] <Iakin3> thelema: Thats Kenmans grahp
[9:22] <thelema> hmm...
[9:41] * TheSeeker (~Fridlekh@cpe-65-28-158-211.socal.rr.com) has joined #Freenet
[10:16] <toad_> hi ppl, i'm back...
[10:16] <thelema> hi toad
[10:16] <toad_> okay
[10:17] <toad_> the root of the major recent problems is probably firstly and foremostly the connections bug
[10:17] <toad_> so i'm going to try to fix it
[10:17] <thelema> did the new connecting code get checked in 8075?
[10:17] <thelema> *5075
[10:18] <toad_> i have no idea
[10:18] * thelema recalls a new connectionOpener
[10:18] <toad_> it's some sort of crypto error
[10:18] <toad_> it must have been in unstable for aaages
[10:18] <toad_> but now it's on both branches
[10:18] <thelema> ok, is there anything I can do to help?
[10:18] <toad_> hmmm
[10:18] <toad_> not yet
[10:18] <toad_> if you'd look into the other issues that'd be useful
[10:19] <toad_> i think the high cpu usage must be caused by the connection bug...
[10:19] <thelema> well, dig in.
[10:19] <thelema> I'll get back into the groove of the binary protocol
[10:20] <toad_> yup, that'd be great
[10:20] <toad_> a little strategic for me to care about right now :)
[10:21] * toad_ digs in like a tank in siege mode...
[10:22] <toad_> (starcraft reference... when they go siege mode they can't move but they do TONS of damage at huge range)
[10:22] <toad_> testbed up, no connections
[10:23] <toad_> Inbound connection failed: freenet.AuthenticationFailedException: Aga
[10:23] <toad_> inst peer (null) @ 127.0.0.1:43404 - Remote does not posess the private key to t
[10:23] <toad_> he public key it offered (terminal) on tcp/connection:
[10:23] <toad_> ...
[10:24] * TheSeeker (~Fridlekh@cpe-65-28-158-211.socal.rr.com) Quit (Read error: 60 (Operation timed out))
[10:28] <thelema> Iakin: looking at what you did with FNCRawMessage, I'm thinking it's *not* the way to go.
[10:38] <toad_> thelema: you'll need hooks for versioned ultra-compact sub-fieldsets e.g. the StoreData fieldsets
[10:39] <thelema> I'm no sure what you're planning as far as versioning, but I see no problems in my design for the StoreData fieldsets.
[10:40] <toad_> except that it won't be very compact - it'd be Estimators.pDNF.1.Center=<bigint> -> 1 byte for estimators, 1 byte for pDNF, 1 byte for 1, 1 byte for Center, the BigInt, some formatting...
[10:40] <toad_> and I doubt it would be 1 byte anyway
[10:40] <toad_> it'd probably overflow if you do it THAT way
[10:42] <thelema> ah, I see what you're trying to do.
[10:42] <thelema> no, I'm not breaking it like that, I'm breaking it even bigger chunks.
[10:42] <toad_> hmm?
[10:42] <thelema> 1 byte for "Estimators.pDNF.1.Center" and then 23 bytes for the bigint.
[10:42] <toad_> ahhh
[10:43] <toad_> except that there's no way that will fit in 1 byte
[10:43] <thelema> Maybe I can make the estimator data two bytes, so we can have up to 255 points in an estimator
[10:43] <thelema> why not?
[10:43] <thelema> why won't it fit in one byte?
[10:43] <toad_> because there's too many of them
[10:43] <thelema> left, center, ra = 3
[10:43] <thelema> * 16 points
[10:43] <thelema> * how many SBKEs?
[10:44] <toad_> Estimators.{pDNF|tDNF|rTransferRate|rTransferFailed|pTransferFailed}.[1-16].{Center|Value (subset!)}
[10:44] <toad_> Value is a subset so you multiply by the size of that subset again
[10:45] <Iakin> [16:44] <thelema> Iakin: looking at what you did with FNCRawMessage, I'm thinking it's *not* the way to go.
[10:45] <Iakin> It might be true..
[10:46] <Iakin> I became a little bit regretful when I saw the result
[10:46] <Iakin> But.. we really need something that is properly supported by a compiler
[10:46] <Iakin> And something that is effective
[10:47] <thelema> Iakin: effective is a hard-coded array.
[10:47] <thelema> toad_: doh. I guess I can allow for the 1-16 to be a byte by itself
[10:47] <thelema> how big is Value?
[10:47] <toad_> then it'd probably be 3 bytes...
[10:47] <toad_> ummm
[10:48] <toad_> either SimpleBinaryRunningAverage.toFieldSet... or BootstrappingDecayingRunningAverage.toFieldSet
[10:48] <thelema> toad_: anyway, how's the crypto going?
[10:48] <toad_> thelema: still tracking...
[10:48] <thelema> is this distracting you from that?
[10:48] * piranha (~piranha@192.220.223.19) has joined #freenet
[10:50] <thelema> Iakin: in the end, a hard-coded array will provide the best controlability for the compression table.
[10:50] <thelema> toad_: 5 estimators. weren't you talking about getting rid of two of them?
[10:52] <toad_> thelema: only if we do a content reset and fix the file size and implement transparent FEC in fproxy
[10:52] * tyesinclair (~tyesincla@170.252.64.1) has joined #freenet
[10:52] <Iakin> thelema: Right.. but I dont like that we force things to serialize themselves, just to _not_ use the serialized value
[10:52] <toad_> we could maybe get rid of one of them
[10:52] * tyesinclair (~tyesincla@170.252.64.1) has left #freenet
[10:52] <toad_> with some changes to how we handle transfer faileds
[10:52] <iip_i2p> <Newsbyte> So, thelema/toad, read my latest post on the maillist? :-)
[10:53] <iip_i2p> <Newsbyte> my devasting, I say devastating, post, toad ;-)
[10:53] <iip_i2p> <Newsbyte> /me wonders if he's going to see another remark from Ian
[10:53] <toad_> Newsbyte: I'm busy fixing the connections bug
[10:53] <Iakin> [16:59] <toad_> except that there's no way that will fit in 1 byte
[10:53] <Iakin> It sure is :)
[10:53] <iip_i2p> <Newsbyte> seriously though; it's clear there are some major probs, aparently missed in unstable
[10:53] <thelema> toad_: oh well. n/m then about less estimators
[10:53] <Iakin> He is skipping the transfer of the decompression table/symbols
[10:54] <thelema> Iakin: that's all you're complaining about in your sedign?
[10:54] <greycat> You'll be happy to hear that "java -cp .../freenet.jar freenet.client.cli.Main put CHK@ some.big.file" works now in stable.
[10:54] <toad_> Newsbyte: the connection bug
[10:54] <toad_> the connection bug
[10:54] <toad_> and the connection bug
[10:54] <toad_> which i am fixing
[10:54] <toad_> there are loads of other things to look at LATER
[10:55] <iip_i2p> <Newsbyte> hmm...so it's all the fault of that bug, is it? ;-)
[10:55] <toad_> much of it
[10:55] <toad_> the cpu usage, the low connections, the consequent crappy routing
[10:55] <toad_> there are other problems but this is the most urgent
[10:55] <iip_i2p> <Newsbyte> so, how comes it wasn't known before, on unstable?
[10:56] <iip_i2p> <Newsbyte> would a real testnetwork instead of current unstable have provided indications what was wrong?
[10:57] <toad_> no
[10:57] <toad_> it would not
[10:57] <toad_> well...
[10:58] <toad_> I suppose instantly upgrading the whole network would have found it pretty fast
[10:58] <toad_> and I am now hunting it in a local testnet
[10:59] <iip_i2p> <Newsbyte> speaking of which...can anyone provide some temporary webspace for hosting (at least the non-wiki) freenethelpsite for some days?
[11:00] <iip_i2p> <Newsbyte> and no, I can't access freenetproject cvs thingy with this puter
[11:02] <toad_> okay, found it
[11:02] <Iakin> [17:10] <thelema> Iakin: that's all you're complaining about in your sedign?
[11:02] <Iakin> Yea.. sort of I guess
[11:02] <toad_> Newsbyte: why can't you get to the freenetproject cvs thingy?
[11:03] <iip_i2p> <Newsbyte> because all my stuff was on the other puter
[11:03] <thelema> Iakin: my current reason for disliking it is that the addition of a backwards-compatible optional field into the protocol kills the compression tables.
[11:03] <iip_i2p> <Newsbyte> cvs client and stuff
[11:04] <toad_> okay, i think i've got it...
[11:04] <iip_i2p> <Newsbyte> it's just temporary, because of my house-rebuilding I talked you about earlier, toad
[11:04] <Iakin> Mmm... thats bad
[11:04] <thelema> toad_: what was/is it?
[11:05] <iip_i2p> <Newsbyte> but I'll have to close down my server for a few days, soon
[11:06] <toad_> hmmm
[11:06] <toad_> who took the .cvsignore out for lib/ /
[11:06] <toad_> ?
[11:06] <Iakin> Not me.... as far as I know
[11:07] <thelema> I didn't even realize it was gone.
[11:07] <toad_> maybe it isn't
[11:07] <toad_> maybe eclipse is just messing me about...
[11:11] <toad_> hmmm
[11:11] <toad_> got one, now it's stuck on a different one?
[11:12] <Pascal> "ant distclean" removes the lib dir
[11:15] <iip_i2p> <Newsbyte> so..ermm...anyone?
[11:15] <thelema> news: sorry; I can't help you out.
[11:15] <toad_> ah, a bug in the bugfix...
[11:15] <toad_> Pascal: sure, the problem is that eclipse wants to add lib/ to cvs
[11:16] <thelema> toad_: I'm still dying to know what the problem was...
[11:16] <toad_> thelema: well, I changed a constructor from new BigInteger(<bytes>) to new BigInteger(1, <bytes>)
[11:16] <toad_> seems harmless enough
[11:16] <toad_> but i think that that change caused it
[11:17] <toad_> have a look at freenet.crypt.Util.MPIBytes(BigInt)
[11:17] <toad_> it's a monster
[11:17] <toad_> not very long but still monstrous
[11:17] <thelema> That would explain why it wasn't caught in unstable; that code was barely checked in when it got merged to stable.
[11:17] <toad_> :)
[11:17] <thelema> I know what that function should be doing...
[11:18] <thelema> I had to re-implement it for ethel
[11:18] <toad_> sh*t, it's still broken...
[11:19] * jay (jcl@ool-18bf6dac.dyn.optonline.net) has joined #freenet
[11:19] <thelema> what's wrong with it?
[11:19] <thelema> It copies in the cyted from the bigint
[11:19] <thelema> bytes
[11:19] <thelema> and then it prepends the size bytes
[11:20] <thelema> and there's no BigInt constructor
[11:20] <toad_> there's an easier and far less problematic way to do it
[11:20] <toad_> mostly the problem is that it calculates the bitlength wrong for the new constructor
[11:20] <thelema> well, FWIW, MPI is an actual standard.
[11:20] <thelema> new constructor?
[11:21] <thelema> num.bitLength() doesn't work on NBIs?
[11:21] <toad_> no, that works fine...
[11:21] <toad_> i dunno, that's my best guess
[11:22] <toad_> i've rewritten it to be more straightforward
[11:22] <thelema> more straightforward? The functions I'm seeing are quite straightforward.
[11:22] <thelema> even if they don't use temp variables to store things.
[11:22] <iip_i2p> <Newsbyte> it could just be put on a homepage, as an orphaned site, and I'll redirect to it with the domainname
[11:26] <toad_> hmmm
[11:26] <toad_> thelema: what exactly is the value in [0] and [1] ?
[11:26] <toad_> is it the num.bitCount() ?
[11:27] <thelema> yes.
[11:27] <thelema> the high 8 bits in [0] and the low 8 bits in [1]
[11:27] <toad_> or is it bitLength()
[11:27] <toad_> ?
[11:27] <thelema> bitlength()
[11:27] <thelema> what's the difference?
[11:27] <toad_> hmmm
[11:27] <toad_> should be the same thing for positive ints...
[11:28] <toad_> no, we'll get off-by-1-errors
[11:28] <toad_> because num.toByteArray() is... argh
[11:28] <toad_> that's what all that stuff was for :|
[11:28] <toad_> hrrm
[11:29] <toad_> BigInteger.toByteArray() is MSB, right?
[11:29] <toad_> most significant byte is in [0] ?
[11:29] <thelema> yes.
[11:29] <thelema> it's big endian and msb is [0]
[11:29] <thelema> Returns a byte array containing the two's-complement
[11:29] <thelema> representation of this BigInteger. The byte array will be in
[11:29] <thelema> big-endian byte-order: the most significant byte is in
[11:29] <thelema> the zeroth element. The array will contain the minimum number
[11:29] <thelema> of bytes required to represent this BigInteger, including at
[11:29] <thelema> least one sign bit, which is
[11:30] <thelema> (ceil((this.bitLength() +
[11:30] <thelema> 1)/8)). (This representation is compatible with the
[11:30] <thelema> (byte[]) constructor.)
[11:30] <toad_> so if it's divisible by 8, then [0] will be == 0
[11:31] <thelema> MPI requires always one sign bit at the beginning of the value.
[11:32] <jay> toad_: did you actually sleep last night?
[11:32] <toad_> not if you use new BigInteger(0, <bytes>)
[11:33] <toad_> jay: yes, it's not that I have trouble sleeping, it's going to bed, going to sleep that is irritating...
[11:33] <toad_> okay, got one connection....
[11:33] <toad_> still only one, which is rather odd...
[11:34] <jay> i see toad's comments all through the log without interruption.. like he napped 30 minutes every couple of hours
[11:34] <thelema> does the problem go away if you replace the new NBI(1,x) with new NBI(x)?
[11:34] <toad_> probably but I don't see why we should do that
[11:34] <toad_> jay: you don't have timestamps
[11:34] <toad_> the IRC goes really quiet through the night
[11:35] <jay> toad_: the irc-log i looked at... not too closely i suppose
[11:35] <thelema> toad_: to make sure that it's the problem.
[11:35] <jay> it only gets real quiet from around 4am-9am EST
[11:36] <jay> which is 8:00-13:00 UTC
[11:37] * thelema is tempted to write a testMPI() function to generate random bigints and start pushing them through writeMPI and readMPI repeatedly, making sure they come out right.
[11:37] <toad_> hmmm
[11:37] <thelema> if I had eclipse (or even a java environment) on this computer, I would.
[11:37] <toad_> it's getting a few connections
[11:38] <toad_> but the overwhelming majority are failing
[11:38] <thelema> what check fails to make one node think the other doesn't have the private key?
[11:38] <jay> im getting repeated freenet.node.rt.OptimizingSlidingBucketsKeyspaceEstimator errors: Guessed probability: 1.0008847182908736 ...
[11:39] <thelema> hmm...
[11:39] <thelema> jay: what do your estimators look like for that node?
[11:40] <jay> thelema: how would i check ?
[11:40] <thelema> in the node Reference Status, click on the node's tcp address for lots of details.
[11:40] <jay> i see
[11:41] <jay> im not sure what node is generating those log entries
[11:41] <jay> looks like lots of them are
[11:41] <thelema> those details would include whole serialized estimators
[11:41] <thelema> then go ahead and reach deep into a random node or two.
[11:42] * TheSeeker (~Fridlekh@cpe-65-28-158-211.socal.rr.com) has joined #Freenet
[11:42] <jay> i have lots of bad node entries in the rt i believe
[11:42] <thelema> bad node?
[11:42] * TLF (francisco@217.Red-81-40-113.pooles.rima-tde.net) has joined #freenet
[11:43] <jay> yeah like void/(void)
[11:43] <thelema> oh, no return information. How many?
[11:44] <jay> i have 20 void/void entries
[11:44] <jay> out of 278
[11:45] * SkyNet (~linuxbazi@www.php-nuke.hu) has joined #freenet
[11:45] <SkyNet> Hi
[11:45] <Iakin> [17:53] * thelema is tempted to write a testMPI() function to generate random bigints and start pushing them through writeMPI and readMPI repeatedly, making sure they come out right.
[11:45] <Iakin> Good idea
[11:45] <thelema> that's a bit high, but 278 is a lot of nodes...
[11:46] <toad_> <thelema> what check fails to make one node think the other doesn't have the private key? - signature verification in FnpLink
[11:46] * SkyNet (~linuxbazi@www.php-nuke.hu) has left #freenet
[11:46] <jay> thelema: does that cleannodes script help in this context at all?
[11:48] <jay> really i don't know how to interpret all the data on the reference status page
[11:48] <jay> many of my estimates are NaN
[11:48] <thelema> jay: are you looking at the details for one node?
[11:49] <jay> both the reference page and details for some nodes that look interesting
[11:49] <jay> im going back and forth between the details and the full list
[11:49] <thelema> on the details, you should see bucket information; leftEdge positions
[11:51] <toad_> ugh
[11:51] <toad_> UGH UGH UGH UGH UGH
[11:51] <jay> the only place the word 'bucket' appears is erTransferFailed.Implementation=SlidingBucketsKeyspaceEstimator
[11:51] <toad_> another f*cking NIO bug...
[11:52] <thelema> and for erTransferFailed, do you see "DividingKey"?
[11:52] <toad_> I hate NIO bugs
[11:52] <toad_> especially the "there's a 50% chance it's in sun's code" bit
[11:52] <thelema> toad_: the other thing this could have been caused by is the new padding code.
[11:52] <jay> DividingKey=e000000000000000000000000000000000000000000001
[11:53] <toad_> thelema: I don't see how, but of course if the trail leads there...
[11:53] <thelema> jay: there should be 16 of them.
[11:53] <jay> this from the page for tcp/218.115.232.122:3456
[11:53] <toad_> anyway, what's happening:
[11:53] <toad_> I fixed some bugs in crypt/, which I recently introduced
[11:53] <toad_> quite subtle nasty ones involving bitlengths and bytelengths
[11:53] <thelema> toad_: I really liked how you implemented the padding. How does it work if the message is almost n*256 bytes long; i.e. a void message won't fit?
[11:53] <toad_> now the node connects to a few nodes
[11:53] <toad_> thelema: it goes to the next block size
[11:54] <toad_> but void is only 4 bytes
[11:54] <jay> thelema: ah i see.. which one am i concerned with ?
[11:54] <thelema> toad_: oh.
[11:54] <toad_> but the vast majority of connections STILL fail
[11:54] <thelema> jay: you want to make sure that the values are increasing.
[11:54] <jay> from Points.1.Value up to Points.a.Value, ...
[11:54] <toad_> this appears to be because of NIO reporting that a connection is readable, then reading zero bytes from it, then not reporting that it is readable ever again
[11:55] <toad_> eventually it times out...
[11:55] <toad_> :<
[11:55] <thelema> jay: from points.1.DividingKey to points.f.DividingKey
[11:55] <toad_> I'm pretty sure the data has been written
[11:55] <toad_> we need to reinstate fixKeys()
[11:55] <Iakin> toad: That I can tell you the reasond for
[11:55] <toad_> I can't believe anyone in their sound mind would use java.nio ...
[11:55] <toad_> Iakin: hmm?
[11:55] <Iakin> It is because we dont clear the selectedSet
[11:55] <thelema> toad_: fixKeys()?
[11:55] <jay> toad_: me neither
[11:56] <Iakin> Which the code _must_ do
[11:56] <toad_> Iakin: I thought that was supposed to be read-only?
[11:56] <toad_> that's what the docs say iirc
[11:56] <Iakin> Nope.. just its iterator
[11:56] <Iakin> not the set itself
[11:56] <thelema> jay: the values are allowed to be pretty much anything (although if you have a value > 1 or < 0 for a probability estimator, that's bad.
[11:56] <Iakin> After a select the code should process all keys in selectedSet then call selectedSet.clear()
[11:57] <Iakin> Before doing the next select()
[11:57] <jay> brb
[11:57] <toad_> so what's happening? We get a spurious report, that adds the key, then we don't remove it, so sel.selectNow() doesn't check whether it is ready at all, and just reports that it isn't?
[11:58] <toad_> I don't get it
[11:58] <toad_> it reported that it was last time...
[11:58] <toad_> but it wasn't
[11:59] <Iakin> That is because the previous loop-spin consumed the 'writability'
[11:59] <toad_> Keys ready: <> is just the size of selectedKeys anyway
[11:59] <toad_> hmmm
[11:59] <toad_> so the previous spin wiped the writability...
[11:59] <Iakin> _nothing_ is automatically removed from the selectedSet
[11:59] <toad_> but it DIDN'T check the writability?
[11:59] <toad_> Iakin: there are zero elements in the selectedSet
[11:59] <toad_> brb
[12:00] <KenMan> 5086 "eighty-sixed" the code. And 5087 would have the chance to add better floating point...
[12:00] <Iakin> Soo? The select timed out then.. or what?
[12:01] <toad_> Iakin: it's a nice theory but I can't make it fit the facts...
[12:01] <toad_> Keys ready before fixKeys: 0/14
[12:01] <toad_> Keys ready: 0/14
[12:01] <Iakin> Mmm?
[12:01] <Iakin> Soo
[12:01] <toad_> so there's nothing in the selectedSet
[12:02] <Iakin> does fixKEys work on the selectedSet?
[12:02] <Iakin> Or on another set?
[12:02] <Iakin> Like.. say, just any set.. with the name selectedSet
[12:03] <toad_> Keys ready before fixKeys is simply currentlyActive ... oh
[12:03] <toad_> it's the return value
[12:03] <toad_> it's not selectedset size
[12:03] <toad_> okay
[12:03] <toad_> so if i add a clear on it just before select, it'll work?
[12:03] <Iakin> Well.. it would look better if you added it right _after_ the processKeys() code
[12:04] <toad_> sel.selectedKeys().clear(); at the beginning of mySelect() ?
[12:04] <toad_> well that's going to be a little hit and miss, I'll keep it at the beginning for clarity
[12:04] <toad_> Iakin: where is this vital piece of info mentioned?
[12:04] <Iakin> Heh.. /me actually has a 'sel.selectedKeys().clear();' call in his loop
[12:05] <Iakin> The nio book for instance
[12:05] <toad_> where in the nio book?
[12:05] <Iakin> The only thing select() does is to add things to selectedSet
[12:06] <toad_> well i'm glad somebody understands it ;)
[12:06] <toad_> thanks
[12:06] <wind789> are there any flow charts that represent the internal workings of freenet?
[12:06] <toad_> wind789: there are a few...
[12:06] <toad_> some in the papers, some on freenet sites...
[12:06] <wind789> cool
[12:06] <KenMan> probably the sewage system of London would be this best graphic
[12:06] <wind789> lol
[12:07] <Iakin> Good question.. I cannot even find the book
[12:07] <Iakin> w8
[12:08] * jmercier (~jmercier@4.10.164.111) Quit (Read error: 104 (Connection reset by peer))
[12:09] <wind789> has anybody ever considered some kinda freenet add-on that shows diagnostic values changing in real time?
[12:09] <wind789> like a live view instead of snapshots
[12:10] <thelema> wind789: I've considered it, but I don't have the necessary experience with Java GUI programming to make it happen.
[12:11] <wind789> I see
[12:11] <toad_> hmmm
[12:11] <toad_> it STILL doesn't work
[12:11] <thelema> not to mention, it wouldn't be quite real-time, it'd be just quick snapshots (maybe 1 per second), which fred wouldn't be able to sustain; it takes much work to snapshot the routing table.
[12:11] <wind789> I like freenet but I have little programming skill, maybe some day when I learn more I'll be able to help out
[12:11] <wind789> I see
[12:12] <thelema> and while it's being analyzed, it has to be locked so that it won't be modified, and that keeps the rest of the code from using it.
[12:13] <wind789> sounds like a lossing battle
[12:13] <KenMan> and for all the data we already capture, we should be able to analyze it and do corrections, eventually :)
[12:14] <Iakin> bbl
[12:15] <wind789> is it possioble to know the rate of fluctuation for all the data that is captured?
[12:15] <toad_> well, it still breaks
[12:15] <KenMan> crypto ?
[12:15] <toad_> and it's probably an NIO problem as both sides seem to time out
[12:16] <thelema> wind789: rate of fluctuation? How do you measure that for a routing table?
[12:16] <wind789> hehe, I don't know
[12:18] <toad_> blah, how come none of this showed up on unstable?
[12:18] <toad_> I suppose I know the answer: I am singly and entirely responsible for all testing on freenet. nobody else uses it. any apparent users are figments of my imagination and therefore unreliable
[12:18] <KenMan> right on !
[12:19] <KenMan> I've had that knawing feeling at times...
[12:20] <thelema> toad_: there wasn't much of a delay between checking in (1) the padding code and (2) the sign changes to BigInt.
[12:20] <thelema> and merging thenm to stable.
[12:20] <toad_> okay, it's not an NIO problem...
[12:20] <toad_> it's back to wierd crypto bugs
[12:21] <toad_> Trying to read 1708 bytes
[12:21] <thelema> probably, most people thought it was just a minor problem with their nodes.
[12:21] <thelema> toad_: test red/writeMPI()
[12:22] <thelema> try reverting the BigInt sign patchess
[12:22] <thelema> troubleshoot by finding where the problem is.
[12:23] <toad_> byte[] buf = num.toByteArray();
[12:23] <toad_> int len = buf.length;
[12:23] <toad_> if(buf[0] == 0) len--;
[12:23] <toad_> byte[] bytes = new byte[len + 2];
[12:23] <toad_> System.arraycopy(buf, buf.length-len, bytes, 2, len);
[12:23] <toad_> int blen = num.bitLength();
[12:23] <toad_> bytes[0] = (byte) (blen >> 8);
[12:23] <toad_> bytes[1] = (byte) blen;
[12:23] <toad_> return bytes;
[12:24] <toad_> what is wrong with this code?
[12:24] <toad_> that has to be where the error is
[12:24] <thelema> the len--
[12:24] <thelema> the bitLength()
[12:24] <toad_> eh?
[12:24] <toad_> explain please
[12:25] <thelema> the idea is to get an arbitrary-sized integer across a communications channel.
[12:25] <thelema> two bytes give the length of what follows
[12:25] <toad_> the other end has to figure out how many bytes to read based on the bit length we send
[12:25] <toad_> yes, the two bytes give the length in bits :|
[12:26] <toad_> so are you saying that blen should be bitLength() + 1?
[12:26] <toad_> and then take the len-- out?
[12:26] <thelema> definitely take out the len--
[12:26] <toad_> no, with the len-- and not the bitlength, we are definitely broken
[12:26] <toad_> if length % 8
[12:26] <toad_> === 0
[12:27] <toad_> the len-- is to take out the sign byte if length is divisible by 8
[12:28] <KenMan> but you allocate a byte even if the length is zero :(
[12:28] <thelema> KenMan: how many bigints have length 0?
[12:28] <thelema> KenMan: we're talking about crypto values.
[12:28] <thelema> KenMan: that's not a relevant case.
[12:29] <KenMan> i shut up now, i know not the details of bigInt
[12:29] <thelema> keep the sign byte.
[12:29] <toad_> why?
[12:29] <thelema> and use bitLength() + 1
[12:29] <toad_> the other side doesn't expect one, that's the problem
[12:29] <toad_> okay
[12:29] <thelema> the other side will happily use a sign byte if one is present.
[12:30] <thelema> I don't see a problem with the original code.
[12:30] <thelema> and We should just test this code to see if it works.
[12:30] <toad_> except that the original code didn't work
[12:31] <thelema> it didn't?
[12:31] <toad_> yes, it didn't
[12:31] <toad_> it generated those errors
[12:31] <toad_> remember?
[12:31] <thelema> which errors?
[12:31] <toad_> "Remote does not posess the private key to the public key it offered"
[12:31] <thelema> how/why?
[12:32] <thelema> just write a short loop that writes then reads a random biginteger from size 10 bits to size 1000 bits
[12:33] <toad_> thelema: it was writing 1 byte too few
[12:33] <thelema> which byte?
[12:33] <Iakin> toad: page 144 in the nio book
[12:33] <Iakin> Read #4 there
[12:34] <toad_> the last one
[12:34] <thelema> it wasn't writing the least significant byte?
[12:34] <toad_> yup
[12:35] <toad_> somehow it left it out
[12:35] <thelema> really?
[12:35] <thelema> how do you know that?
[12:36] <toad_> well maybe it was the fault of the read sub rather than the write one
[12:37] <toad_> but somebody was messing up
[12:37] <toad_> and a byte was left off the end
[12:37] <thelema> was it only erroring 1 in 8 times?
[12:37] <toad_> no, it was always erroring
[12:37] <thelema> how was it doing it non-consistently?
[12:37] <thelema> how did it work sometimes?
[12:37] <toad_> no, it was doing it consistently
[12:37] <toad_> it didn't until i fixed it!
[12:37] <toad_> it ALWAYS failed
[12:37] <toad_> now it occasionally works
[12:38] <thelema> so fred's crypto has been completely broken until now?
[12:38] <toad_> that's PROGRESS
[12:38] <toad_> no, it's something to do with the bigint changes
[12:38] <thelema> hmm...
[12:38] <toad_> but I'm not going to revert them because they're the Right Way to do it
[12:38] * KenMan likes to see a confident toad
[12:39] <thelema> I'm not so sure about that; it seems really reasonable to just make a few things NBI and leave the rest BI.
[12:39] <toad_> no, not that
[12:39] <toad_> the other one
[12:39] <toad_> these happen on stable too remember?
[12:39] <thelema> what other one?
[12:39] <toad_> in any case we get a factor of ~ 10 gain in speed
[12:39] <thelema> the (1,x) changes?
[12:39] <toad_> on modPow
[12:39] <toad_> which IS used many times in session negotiation
[12:39] <toad_> yeah
[12:40] <thelema> hmm...
[12:40] <thelema> well, I'd start out by fixing read/writeMPI fully. And it'll be hard to know whether they're fixed unless you write a testCase.
[12:41] <toad_> well, the length appears to be wrong
[12:41] <toad_> as read
[12:41] <toad_> 4,993 bytes is crazy
[12:41] <toad_> none of our numbers are THAT long
[12:42] <thelema> 4993 bits
[12:42] <toad_> no, 4,993 bytes
[12:42] <toad_> that's what it's reading
[12:42] <thelema> oh my.
[12:42] <thelema> this is why you need a simplified test Case.
[12:42] <toad_> int blen = num.bitLength() + 1;
[12:42] <toad_> bytes[0] = (byte) (blen >> 8);
[12:42] <toad_> bytes[1] = (byte) blen;
[12:42] <toad_> that can't be wrong
[12:42] <toad_> it wants it in bits, and it wants it big endian
[12:43] <thelema> it'll be hell to debug if you don't know what the input was on the other end..
[12:44] <toad_> I log much of that
[12:44] <thelema> you log MPI inputs?
[12:44] * thelema wishes he had a JDK available.
[12:44] <toad_> thelema: what is your f*cking problem? why can't you install a jdk?
[12:45] <toad_> don't you have a shell account somewhere?
[12:45] * Ash-Foxeus (Hal-9000@aaq112.neoplus.adsl.tpnet.pl) Quit (Client Quit)
[12:46] <thelema> I just realized I do, but I'm going to have to develop over a modem connection.
[12:47] <toad_> hmmm
[12:47] <toad_> yucj
[12:47] <toad_> yuck
[12:47] <toad_> it's not +1
[12:48] * toad_ is tempted to rewrite the two methods to send the length in bytes
[12:48] * TLF (francisco@217.Red-81-40-113.pooles.rima-tde.net) Quit ("http://www.vitaliano.esp.cc || Adi?s a todos y a todas y que les vaya bien.")
[12:48] <toad_> unfortunately that would break backwards compatibility
[12:48] <thelema> >>3 or <<3
[12:48] * Ash-Fox (Hal-9000@aaq112.neoplus.adsl.tpnet.pl) has joined #FREENET
[12:48] <toad_> byte[] data = new byte[(((b1 << 8) + b2) + 8) >> 3];
[12:48] <toad_> that expects bitLength()
[12:49] <toad_> and it expects it to be padded with a sign byte
[12:49] <toad_> it would be much clearer if we could just send the length in bytes
[12:49] <thelema> ok, so what's the problem?
[12:53] <toad_> okay, the first 8 conn attempts fail... normal, because we've just started up, the other nodes may not be up yet
[12:53] <toad_> then 2 succeed
[12:54] <toad_> then there are no more attempts... wtf?
[12:55] <Iakin> toad: also read 4.3.3 in da book
[12:56] <toad_> hmmm
[12:56] <toad_> another 7 attempts, all failed
[12:59] * toad_ reverts to the cvs version...
[12:59] <toad_> of MPIbytes
[13:00] <Iakin> 'Selections are cumulative. Once a selector adds a key to the selected set, it never removes it. And once a key is in the selected set, ready indications in the ready set of thay key are set but never cleared.'
[13:01] <toad_> hmmm
[13:01] <toad_> it appears to be in NBI
[13:01] <toad_> but then why does it happen on stable?
[13:01] <Iakin> '.. assigns responisibility to the programmer to properly manage the keys to ensure that the state information they represent does not become stale.'
[13:02] <toad_> aha, the byteValue() caching...
[13:05] <Iakin> Hmm.. what up that that one?
[13:05] <toad_> if i had found that immediately, i wouldn't have fixed the nio bug :)
[13:05] <toad_> Iakin: we cache byteValue() in NBI
[13:05] <toad_> except that it's not necessarily the right length
[13:05] <Iakin> Yes, I worte that code
[13:05] <toad_> if it's fed in in the constructor, it may be 1 byte too short
[13:05] <Iakin> Ahhhhh
[13:06] <toad_> okay, fixed, all gone
[13:06] <toad_> it took a while, but it's GONE !!
[13:06] <Iakin> So.. dont assign it in the constructor...
[13:06] <Iakin> Damn..
[13:06] <Iakin> Good catch
[13:06] <Iakin> commit it plz
[13:06] <toad_> no, pitifully poor actually... I ran around chasing my tail for ages while thelema constantly pointed out that I was barking up the wrong street
[13:07] <thelema> :) sometimes I'm right; sometimes I'm wrong. This time I got lucky.
[13:08] <toad_> :)
[13:08] * toad_ updates...
[13:08] <toad_> have to update before committing...
[13:09] * toad_ rebuilds, retests...
[13:10] <toad_> note to self: before committing stable builds verify that it does SOMETHING on testnet
[13:10] <thelema> toad_: that seems reasonable.
[13:11] <toad_> slightly tricky to do with the actual stable code
[13:11] <toad_> but at least verify the unstable code on which it is based
[13:11] <toad_> CANNOT RENAME NODE FILE node-temp TO node
[13:11] <toad_> hmmm
[13:14] <thelema> no, verify the stable code before building snapshots
[13:16] <toad_> Iakin: give me a comment for before the sel.selectedKeys.clear() in mySelect() ?
[13:18] <KenMan> just letting 5086 run for a while has seemed to let it settle down a bit. It took 2 days and a few restarts, but it looks acceptable now.
[13:19] <KenMan> my primary concern was the wobbly number of queries, but they are settling down into the 1K-2K/hr range .
[13:20] <toad_> thelema: well normally I verify the stable code by running a node for a while and testing that
[13:21] <toad_> strangely that wasn't enough in this case
[13:22] <toad_> now, how much should i include in the stable merge?
[13:22] <toad_> just this bugfix?
[13:22] <toad_> or the nativebigint stuff as well?
[13:22] <toad_> actually... stable doesn't have this bug, it doesn't have the nativebigint stuff...
[13:22] <toad_> but it still gets.. hmmm
[13:23] <toad_> might be the nio bug
[13:23] * TheSeeker (~Fridlekh@cpe-65-28-158-211.socal.rr.com) Quit (Read error: 110 (Connection timed out))
[13:24] <toad_> hmmm
[13:24] * toad_ getting a LOT of Interpolated crazy value: -1998.766210074004
[13:24] <toad_> negative interpolated values
[13:24] <KenMan> that can't be good
[13:24] * toad_ hrrms at both Iakin and thelema
[13:25] <toad_> Iakin has of course suddenly left us
[13:25] <toad_> well not so suddenly :|
[13:25] <Iakin> [19:32] <toad_> Iakin: give me a comment for before the sel.selectedKeys.clear() in mySelect() ?
[13:25] <Iakin> He's here..
[13:25] <Iakin> on and off :)
[13:25] <toad_> Iakin: I got one from above
[13:25] <toad_> /**
[13:25] <toad_> * From the NIO book:
[13:25] <toad_> * 'Selections are cumulative. Once a selector adds a key to the
[13:25] <toad_> * selected set, it never removes it. And once a key is in the
[13:25] <toad_> * selected set, ready indications in the ready set of thay key
[13:25] <toad_> * are set but never cleared.'
[13:25] <toad_> * Thus you need to clear the selected set before selecting.
[13:25] <toad_> * Effects of not doing this: Missed notifications.
[13:26] <toad_> */
[13:26] <toad_> Iakin: you broke routing!
[13:26] <toad_> well i think it was you...
[13:26] <Iakin> In what way?
[13:26] <toad_> lots of crazy guessed values
[13:26] <toad_> and also unhandled throwable in Checkpoint: Check for connections to open: java.lang.ArrayIndexOutOfBoundsException: 16
[13:27] <toad_> lots and lots and LOTS of ERRORs
[13:27] <toad_> so now i'm going to have to merge from a non-working unstable branch
[13:27] <toad_> i.e. i'll have to leave nativebigint out, at least out of routing
[13:28] <Iakin> Hmm..interesting
[13:28] <toad_> or I could just revert your commits over the last few days?
[13:29] <toad_> hopefully they're on the cvs list...
[13:29] <thelema> they are...
[13:29] <toad_> java.lang.ArrayIndexOutOfBoundsException: 16
[13:29] <toad_> at freenet.node.rt.BucketStore.findIndex(BucketStore.java:389)
[13:29] <toad_> at freenet.node.rt.SlidingBucketsKeyspaceEstimator.guess(SlidingBucketsK
[13:29] <toad_> eyspaceEstimator.java:312)
[13:29] <toad_> at freenet.node.rt.SlidingBucketsKeyspaceEstimator.guessTime(SlidingBuck
[13:29] <toad_> etsKeyspaceEstimator.java:225)
[13:29] <toad_> at freenet.node.rt.StandardNodeEstimator.longEstimate(StandardNodeEstima
[13:29] <toad_> tor.java:861)
[13:29] <thelema> oops, dropped a %
[13:30] <thelema> Iakin: I told you to watch out for that last %
[13:30] <Iakin> toad: To me it feels more logical to clear the set after processing connections, rather than befor
[13:30] <Iakin> e
[13:30] <Iakin> The *last* % ?
[13:30] <Iakin> or.. _last_
[13:30] <thelema> in the case where idx = 16, you need a % to make sure you're accessing bucket 0
[13:31] <toad_> Iakin: to me it seems more reliable
[13:31] <toad_> to clear before processing
[13:31] <Iakin> toad: Are you completely sure that the code cannot call select from more than one place..
[13:31] <Iakin> because if it does.. boom.. broken
[13:32] <toad_> Iakin: hmm?
[13:32] <toad_> we do a selectNow somewhere
[13:32] <Iakin> boom!
[13:32] <Iakin> :)
[13:32] <toad_> Iakin: ummm
[13:32] <toad_> please don't do what you've done in BucketStore again
[13:32] <toad_> it may be beautiful to you
[13:32] <Iakin> toad: Clear it where appropriate..
[13:33] <toad_> but to me it's horribly wasteful
[13:33] <toad_> and it's on the critical path
[13:33] <toad_> and it's broken everything
[13:33] <Iakin> toad: On and off code does that
[13:33] <toad_> what the fuck is FindResult for?
[13:33] <toad_> for example
[13:33] <Iakin> To be able to understand what happens
[13:33] <Iakin> (for a person that is not the creator of the code)
[13:34] <toad_> I suppose we do a load of object creation in interpolate...
[13:34] <toad_> blah
[13:34] <thelema> toad_: it's a nice way of returning two values in one return; bucketvalue, direction
[13:34] <toad_> well, if you're going to make it more readable, please make it work
[13:34] * toad_ suggests you make a branch if you're going to commit MAJOR untested code
[13:34] <thelema> In a language with good data structures, I couldn't hesitate to use something like FindResult.
[13:35] <thelema> s/couldn't/wouldn't/
[13:35] <lolo-laptop> hmm... so why's it broken, and why can't we just fix it instead of backing it out?
[13:35] <Iakin> toad: The reason I couln't try it what that you had broken routing just before
[13:35] <toad_> okay, FindResult is reasonable
[13:35] <toad_> Iakin: I had?
[13:35] <toad_> what have I done to routing recently?
[13:36] <Iakin> The biginteger stuff..
[13:36] <KenMan> yes but that is Toad's *job* :o
[13:36] <Iakin> My node worked before that
[13:36] <toad_> you mean the connections?
[13:36] <Iakin> Right
[13:36] <toad_> well, please think about it
[13:36] <Iakin> Without connections I couldn't really test it
[13:36] <toad_> I publicly announced that the connections had broke both branches
[13:36] <toad_> that means that when I fixed it I would like to merge the bug fix immediately
[13:36] <toad_> THAT MEANS THAT COMMITTING MAJOR DISRUPTIVE CHANGES WITHOUT ANY TESTING IS A BAD IDEA
[13:37] <toad_> okay?
[13:37] <Iakin> It is a one-liner, no?
[13:37] <thelema> toad_: relax and revert the changes.
[13:37] <thelema> Iakin: don't check in rewrites like this to a broken codebase.
[13:37] <toad_> you're free to commit them to a broken codebase
[13:38] <toad_> except when the brokenness is also on stable
[13:38] <Iakin> ?
[13:38] <Iakin> aha
[13:38] <Iakin> Well, unfortunate circumstances..
[13:38] <toad_> quite
[13:38] <toad_> and now you will revert the changes
[13:39] <toad_> hmmm
[13:39] <toad_> are they all in node/rt/ ?
[13:39] <thelema> yes.
[13:39] * toad_ reverts the changes
[13:39] <thelema> and all you need to do is revert SBKE
[13:39] * KenMan hollers LUNCH-TIME. chicken & steak fajitas, with bacon...
[13:39] <thelema> leave bucketStore alone, and it won't interfere with anything.
[13:40] <Iakin> ok, I'll be more carefull next time
[13:41] <Iakin> ...but.. I was under the impression that you knew what was wrong even last friday or saturday or so
[13:41] <thelema> me too.
[13:42] <Iakin> So, I thought I could move on with the stuff I had prepared for quite some time
[13:42] <toad_> well whatever
[13:42] <toad_> i'm going to commit the reversion soon
[13:43] * shendaras (shendaras@ip68-97-113-208.ok.ok.cox.net) has joined #freenet
[13:44] * TLF (francisco@88.Red-81-40-113.pooles.rima-tde.net) has joined #freenet
[13:44] <toad_> of course, the problem on stable can't possibly be the NBI problem...
[13:44] <toad_> it might be the NIO bug
[13:44] <toad_> (thanks Iakin for finding that)
[13:45] <Iakin> the nio bug has been around for months (or ever even)
[13:45] <Iakin> And I _have_ hinted on it before here ;)
[13:45] <toad_> nah, every so often we do a stack of work on NIO
[13:45] <toad_> and break it differently
[13:45] <toad_> Iakin: :|
[13:46] <Iakin> I didn't really think it was important though..
[13:46] <toad_> well, if it makes us miss reads, it could be quite important
[13:47] <Iakin> I thought that the worst effect would be some CPU wasteage..
[13:51] <toad_> well it's quite possible that there's another big bug causing stable suckage...
[13:51] <toad_> 8 minutes uptime, 18 connections...
[13:53] <wind789> is that on your testnet?
[13:53] <wind789> or unstable?
[13:53] <toad_> cool, new sites on tfe...
[13:53] <toad_> wind789: on unstable
[13:54] <toad_> 22 conns in 10 minutes...
[13:54] <wind789> I will update my unstable
[13:54] * toad_ sets successfulDelayCutoff to 1000ms on unstable, and requestDelayCutoff to 500ms
[13:55] <thelema> 2.2 conns per minute? 25ish seconds per connection?
[13:55] <toad_> thelema: hmmm
[13:55] <toad_> 27 conns in 12 minutes...
[13:55] <toad_> consistent with the above...
[13:56] <wind789> is out of memory(on startup) a known issue with nodes that have tons of DS content?
[13:56] <toad_> hmmm
[13:56] <toad_> Fishman papers were of course only inserted on stable
[13:56] * toad_ needs to write those binary migration tools...
[13:56] <toad_> or at least decide whether they are important
[13:56] <toad_> wind789: it can happen... but if it OOMs on startup, it's not going to start up :|
[13:57] <wind789> indeed
[13:57] <toad_> hmmm
[13:57] <toad_> he did, apparently
[13:57] <toad_> at least part of the site iis on unstable
[13:57] <wind789> I updated my FLaunch.ini to compensate
[13:58] <toad_> well, we DO use a certain amount of memory in proportion to the number of files in the store
[13:58] <toad_> (well, maybe n log n)
[13:59] <toad_> most of it is O(n) memory, some of it is O(n log n) memory
[13:59] <toad_> SEC Jul 27, 2004 7:17:14 PM (freenet.transport.WriteSelectorLoop, Network writing thread, NORMAL): fixKeys added sun.nio.ch.SelectionKeyImpl@13504f6
[13:59] <toad_> hmmm
[13:59] <toad_> something interesting haappening herre...
[13:59] <toad_> Iakin: help?
[14:00] <toad_> Iakin: some interesting NIO happenings...
[14:01] <toad_> hmm, went away
[14:01] <toad_> okay
[14:01] * toad_ will ignore for now
[14:02] <wind789> I've got about 130,000 keys in my DS about 45 gigs
[14:02] <toad_> so 362kB/key
[14:02] <toad_> hmmm
[14:03] <toad_> that's another issue... if we reduce the max file size, we have to improve the native ds's memory usage per key
[14:03] <toad_> we can probably do that using thingy filters though...
[14:04] <thelema> let's work on less radical solutions; fixing NIO and transport layer bugs until everything there is like greased lightning
[14:04] <thelema> kalman filters?
[14:04] <toad_> thelema: for now
[14:05] <toad_> not kalman filters, another kind of filter
[14:05] <thelema> bloom filters?
[14:05] <toad_> yeah
[14:05] <thelema> the compressed bitarrays
[14:06] <toad_> what the..?
[14:06] <toad_> it didn't get committed...
[14:06] <wind789> mine shows 60168 should it be higher?
[14:09] <toad_> no
[14:12] <toad_> I didn't commit
[14:12] <toad_> it
[14:12] <toad_> or eclipse didn't commit it
[14:12] <toad_> or something
[14:12] <wind789> wow I finnaly have 1 contacted node on unstable
[14:12] <wind789> I havn't had that for a few days
[14:12] <toad_> wind789: i got 20 pretty fast with the new code...
[14:12] <wind789> cool
[14:13] <toad_> |This web page presents a declaration filed by Steven Fishman on April 9, 1993 in which he included the OT (Operating Thetan) materials as exhibits.
[14:13] <toad_> YAY!
[14:14] <wind789> lol it went back to 0
[14:14] * Hory (~Miranda@81.196.25.110) has joined #FreeNet
[14:14] * toad_ attempts to commit 60172
[14:14] <toad_> with fixes
[14:14] <toad_> and some other changes
[14:16] * TLF is away: Por ah?, de juerga.
[14:17] * lolo-laptop fires up his unstable freenet on stable network with new code.
[14:20] <KenMan> updated query graph -> http://mywebpages.comcast.net/jkcorson/FreeNet/5086c.png
[14:21] <wind789> restarting node with new jar...
[14:21] <thelema> KenMan: is queries/min incoming or outgoing?
[14:21] <KenMan> it is incoming requests
[14:21] <KenMan> i suppose the graph should say that...
[14:23] * toad_ announces 60172
[14:24] <lolo-laptop> Jul 27, 2004 1:42:29 PM (freenet.transport.WriteSelectorLoop, Network writing thread, NORMAL): fixK
[14:24] <lolo-laptop> blah
[14:25] * toad_ changes topic to 'Freenet: Taming the World's Largest Tamagotchi | Upgrade to 5086 | Unstable: Upgrade to 60172 | Logs: http://newton.matcmp.ncc.edu/~lockej/freenet/chanlog/'
[14:25] <thelema> KenMan: all accepted?
[14:26] <toad_> KenMan: ugh
[14:26] <KenMan> almost. 2:00-4:00 GMT they were .999 .985 and .872 , but otherwise yes.
[14:26] <toad_> that's not good
[14:26] <KenMan> rather, 14:00-16:00 GMT...
[14:26] <toad_> what's your outgoingRequestInterval ?
[14:26] <toad_> does it fluctuate like that?
[14:26] <thelema> well, there's still some resonance in the network...
[14:26] <KenMan> lesse...
[14:27] <toad_> is it very close to 600,000ms most of the time?
[14:27] <toad_> thelema: perhaps
[14:27] <wind789> does it take a few minutes for the update to hit http://freenetproject.org/snapshots/ ?
[14:27] <toad_> wind789: it should be there now for unstable
[14:27] <toad_> but yes
[14:27] <wind789> k
[14:27] <toad_> in fact i have to activate it manually...
[14:27] <toad_> or it takes hours
[14:27] <wind789> hehe k
[14:28] <KenMan> i'm not capturing outgoingMRI/minute yet. But per hour, it is on a general upward trend, which is why things have settled down, i suppose.
[14:29] <toad_> KenMan: what's it look like for the last hour's worth of minutes?
[14:29] <toad_> is it wildly varying?
[14:29] <toad_> what's the approximate value?
[14:29] <KenMan> no, outgoingMRI is around 110s now. Last hour is squiggly...
[14:29] <toad_> hmmm
[14:29] <toad_> how squiggly?
[14:29] <thelema> 2 minutes... ick.
[14:30] <KenMan> actually it is up around 320s now... it peaks quickly, dips LOW, and then peaks again (last 60 mins)
[14:31] <KenMan> it is more useful to examine the MRIs for a single link , but I'll start charting the per minute average diagnostic...
[14:31] <KenMan> i think the per-link values wiggle more than the average, just by the nature of using an average.
[14:32] <KenMan> I'll have to reinstrument the code to log individual values again :(
[14:32] <thelema> slower changes!
[14:33] <KenMan> believe me, from watching incoming mRIs from 1 peer, you can make some educated guesses about what that peer is doing...
[14:33] <KenMan> ( as in, "frosting" )
[14:33] <verl> hm, is native bigint currently in unstable and is supposed to work with the files you get from freenetproject.ord/snapshots? right now it's jcpuid-windows.dll in freenet-ext.jar but freenet is searching for one called jcpuid-x86-windows.dll
[14:34] <toad_> KenMan: the code is plenty instrumented
[14:34] <toad_> you just need to find the instrumentation :)
[14:34] <KenMan> if we get I know, that's the problem - too much data gets logged !! :) I had bad luck getting individual MRIs to write out however...
[14:34] <toad_> verl: it's there
[14:34] <toad_> it should work
[14:35] <toad_> logLevelDetail=freenet.PeerHandler:debug
[14:35] <KenMan> if we get back to balancing local versus remote queries, I think that the local node is only deserving of 1/n of query capacity (n=# peers)
[14:35] <KenMan> for reasons of anonymity, mainly
[14:35] <toad_> hmmm
[14:35] <toad_> my freenet-ext.jar differs from the one on dodo
[14:36] <thelema> KenMan: I like that reasoning; nodes that evade local limiting aren't as anonymous.
[14:37] <KenMan> well, the alternatives is balance remote vs local 1:1 (split in half), or (as current) gobble 100% for local requests...
[14:37] <toad_> verl: hrrm
[14:38] <verl> here's what i get:
[14:38] <verl> ERROR: Resource name [freenet/support/CPUInformation/jcpuid-x86-windows.dll] was not found
[14:38] <verl> WARN: Native CPUID library jcpuid not loaded - using pure java
[14:38] <verl> INFO: Non-optimized native BigInteger library 'jbigi-windows-none.dll' loaded from resource
[14:38] <toad_> okay
[14:38] <toad_> i'll fix
[14:38] <verl> thanks :)
[14:38] <toad_> thanks for finding it
[14:39] <toad_> I was just about to merge that stuff to stable!
[14:39] <thelema> verl: native is still much better that java, even non-optimized native.
[14:39] <verl> well for every stupid thing i say, sometime i have be right :)
[14:39] * Superfan (francisco@29.Red-81-40-113.pooles.rima-tde.net) has joined #freenet
[14:39] <toad_> thelema: is it? benchmarks?
[14:40] <verl> thelema: yes, cpu usage is already better
[14:40] <KenMan> it HAS to be !!
[14:40] <thelema> toad_: how about just some anecdotal evidence...
[14:40] <toad_> thelema: fine, gimme :)
[14:40] <thelema> the gmp people are fanatic about performance.
[14:41] <toad_> verl: fetch the new freenet-ext.jar
[14:41] <toad_> it's been updated
[14:41] <toad_> does it fix that?
[14:41] <thelema> toad_: just this right here: <verl> thelema: yes, cpu usage is already better
[14:41] * TLF (francisco@88.Red-81-40-113.pooles.rima-tde.net) Quit (Nick collision from services.)
[14:41] <shendaras> IIRC, the NativeBigInteger from I2P comes with a simple benchmark to compare...
[14:42] <thelema> cpu optimizations are nice, but excluding SSE/MMX they're not a big deal (<10% usually)
[14:42] <verl> INFO: Native CPUID library jcpuid loaded from resource
[14:42] <verl> INFO: Optimized native BigInteger library 'jbigi-windows-athlon.dll' loaded from resource
[14:42] <verl> !
[14:43] <lolo-laptop> damn, stable is sucking down the bw now that CPU usage is down...
[14:43] <lolo-laptop> 61577 bytes/second (128.3%) (unstable code, stable network)
[14:43] <KenMan> in this context, optimization using assembly (or even just C) can make a big difference, but we are comparing using java to using efficient C code here.
[14:44] <toad_> verl: cool
[14:44] <toad_> brb
[14:44] <KenMan> without GMP, we are using a different implementation, one that may be slower than GMP. Without any fancy CPU optimizations even being considered.
[14:44] <thelema> lolo-laptop: you7r bandwidth limit is > 20K/s
[14:44] <lolo-laptop> In many things, java is just as fast as native C ... it is only when trig and other functions which have specific native implimentations which java hasn't implimented on all the billions of platforms out there yet that java gets slow
[14:44] <lolo-laptop> thelema: what's yer point?
[14:45] <thelema> is it working? most people have problems with >20K
[14:45] <KenMan> i'm not trying to knock the java vs C thing, just the actual software implementations being considered...
[14:45] <lolo-laptop> thelema: ahh... it is only sorta working... so yes, I too have problems ;-)
[14:46] <thelema> toad_: that'd be a good thing to work on.
[14:46] <lolo-laptop> KenMan: my point is that you were being too general, it is only highly architecturaly specific things (things which not all java supporting archs have instructions for) that java is slow on these days
[14:46] <KenMan> damn. another chink in the armor
[14:47] <KenMan> so we have reports of faulty bw limiting from about 5 people now...
[14:47] <toad_> hmmm
[14:47] <lolo-laptop> it isn't so bad that I can't use it...
[14:47] <lolo-laptop> just have to set the limit a bit lower than I want it to run
[14:47] <toad_> lolo-laptop: the point is, it would be really fast, if sun would just open the source
[14:47] <toad_> because then they could use GMP
[14:47] <toad_> :)
[14:47] <lolo-laptop> toad_: this is _very_ true :)
[14:48] <thelema> lolo-laptop: there's only 3-4 "java supporting archs" out there.
[14:48] <thelema> and there's a list of 50+ archs with specific optimizations in gmp
[14:49] <toad_> thelema: hrrm, which ones?
[14:49] * toad_ wonders if we should include code for all the java supported archs
[14:49] <toad_> okay, merge question:
[14:49] <KenMan> maybe i subconsciously AM belittling java (vs C). I suspect unoptimized GMP can beat the java BigInteger implementation.
[14:49] <toad_> I have a patch in unstable to remove the RequestInterval from FNP messages
[14:50] <lolo-laptop> thelema: sparc, sparc64, x86, amd64, ppc, ppc64? And aren't there ports for ARM, and MIPS too?
[14:50] <toad_> this saves ~ 23-34 bytes per message and enables me to cut the padding interval to ~ 160 bytes
[14:50] <toad_> lolo-laptop: GCJ runs on loads
[14:50] <toad_> Kaffe runs on loads
[14:50] <toad_> Sun's official VM isn't particularly portable
[14:50] <toad_> now, that's the pro
[14:50] <lolo-laptop> I see
[14:51] <toad_> the con is: it would break rate limiting when talking to 5084 nodes!
[14:51] <toad_> so... probably not?
[14:51] <thelema> windows, linux, solaris sparc, solaris x86, win ia64, linux ia64
[14:51] <robilad> kaffe has been ported to around 70 platforms, afaict
[14:51] <thelema> from the 1.4.2 download page.
[14:52] <toad_> I replace explicit RequestInterval on the FNP messages with sending one MRIPacketMessage in each packet sent
[14:52] <toad_> the problem is that 5084 doesn't know about MRIPacketMessage's
[14:52] <toad_> so probably want to keep them, right?
[14:52] <thelema> toad_: rock the network.
[14:53] <robilad> and i'm not even couting ppc vs. ppc64 as separate platforms ;)
[14:53] <toad_> thelema: given the disasters on 5085 and 5086, is that a good idea?
[14:53] <thelema> This release of the J2EE 1.4 SDK and the Sun Java System Application
[14:53] <thelema> Server Platform Edition 8 is available for the following platforms:
[14:53] <thelema> Solaris (SPARC and x86)
[14:53] <thelema> Sun Java Desktop System
[14:53] <thelema> Windows 2000
[14:53] <thelema> Windows XP
[14:53] <thelema> Windows Server
[14:54] <toad_> robilad: most of them wierd embedded things, correct?
[14:54] <thelema> Red Hat Enterprise Linux
[14:54] <toad_> having said that...
[14:54] <robilad> toad_: uh, yes ;)
[14:54] * toad_ looks forward to running freenet on his zaurus!
[14:55] <lolo-laptop> toad_: I would leave it out till next merge if I were in your shoes.
[14:55] <thelema> toad_: what about 5085 and 86? get people here to bootstrap the network, and run insert/retrieval tests to see if it works
[14:56] <thelema> toad_: if it does, snapshot and recommend people upgrade.
[14:56] * thelema thinks about it and realizes toad is probably right
[14:56] <robilad> toad: arm, alpha, cris, i386, ia64, m68k, mips, parisc, s390, sh, sparc, x86_64. I know there is a cray port, but i haven't seen the patches yet
[14:57] <Iakin> [20:54] <verl> WARN: Native CPUID library jcpuid not loaded - using pure java
[14:57] <robilad> and each has one or more operating systems ;)
[14:57] * piranha (~piranha@192.220.223.19) Quit (Read error: 60 (Operation timed out))
[14:57] <Iakin> Heh.. navtive cpuid.. sure..
[14:57] <Iakin> I'll fix that tomorrow or so
[14:57] <Iakin> (the log message that is)
[14:57] <toad_> Iakin: it IS native
[14:57] <toad_> that's the problem
[14:58] <robilad> toad_: well, it seems like just a temporary problem for me.
[14:58] <robilad> if we get it to work with classpath's nio
[14:59] <robilad> all you need to do is ship an exe to your window users generated with gcj
[14:59] <robilad> and be done with it
[14:59] <robilad> that's where gcj's potential is :)
[15:00] <robilad> i have no idea how gcj-ed pure java big math fares vs. gmp, though
[15:01] <thelema> robilad: if you can compile the NBI library in GCJ, you can test it.
[15:01] <robilad> nbi?
[15:01] <thelema> NativeBigInteger
[15:01] * robilad goes to fetch some freenet
[15:01] <vsalento> is it on the todo list to use locally compiled libgmp if there is one? I need to replace to libjbigi-linux-none.so with one that uses shared lib since my arch is not supported.
[15:02] <hirvox> vsalento:you need to modify the build.sh
[15:02] <vsalento> I have done so but it would be nice if that wouldn't be neccessary... not everyone knows how to do it
[15:03] <thelema> Iakin: would it be tough to look for a libjbigi-local before doing all the probing?
[15:03] <vsalento> especially since in gentoo my gmp is optimized already for my cpu
[15:04] <hirvox> vsalento:you only need to uncomment one line and comment another in build.sh
[15:04] <Iakin> toad: WARN: Native CPUID library jcpuid not loaded - using pure java
[15:04] <Iakin> There is no native CPUID implementation..
[15:04] <vsalento> of course the freenet ebuild could do all the necessary things...
[15:04] <Iakin> _That_ was what I meant
[15:04] <Iakin> Grr...
[15:04] <Iakin> s/native/java
[15:05] <Iakin> thelema: Nope
[15:05] <vsalento> what happens if you try to use libjbigi that is linked dynamically to libgmp and there isn't one available?
[15:06] <Iakin> vsalento, what arch?
[15:06] <lolo-laptop> Jul 27, 2004 2:23:07 PM (freenet.transport.WriteSelectorLoop, Network writing thread, NORMAL): fixKeys added sun.nio.ch.SelectionKeyImpl@10e791d **whine whine** :-P :-P
[15:06] <hirvox> vsalento:yes, the packager should do it for the users. <plug>for example, my RPMs do optimize the libjbigi for the target architecture</plug>
[15:06] <vsalento> VIA Eden... not sure what it actually is but some i586 with some extras
[15:06] <Iakin> vsalento: It will fallback..
[15:07] <Iakin> To java code..
[15:07] <Iakin> (since it will fail loading the .so that requires the *other* .so)
[15:08] * piranha (~piranha@192.220.223.5) has joined #freenet
[15:08] <Iakin> vsalento, is there a target for it in gmp?
[15:08] <vsalento> can you first try to load .so that is dynamically linked to libgmp and then try those arch-specific .so? and it both fails then fallback to java impl?
[15:09] <thelema> vsalento: that was what I suggested, more or less
[15:09] <vsalento> I don't know... I just ran 'emerge gmp' and then when dynamically linked libjbigi to that the speed was 5 times better than using libjbigi-linux-none.so
[15:09] * Ash-Fox (Hal-9000@aaq112.neoplus.adsl.tpnet.pl) Quit ("Leaving")
[15:09] <Iakin> Current sequence: 1. choose cpu-specific or generic (the 'none')
[15:09] <thelema> vsalento: really? wow
[15:09] <Iakin> 2. Look in the classpath for the lib choosen above
[15:10] <Iakin> 3. look in the LD_LIBRATRY_PATH
[15:10] * piranha (~piranha@192.220.223.5) Quit (Client Quit)
[15:10] * piranha (piranha@3ffe:b80:1ca1:0:0:0:deca:fbad) has joined #freenet
[15:11] <vsalento> freenet unstable uptime 43min -> log file 101M wtf?
[15:12] * Ash-Fox (Hal-9000@aaq112.neoplus.adsl.tpnet.pl) has joined #FREENET
[15:12] <greycat> what verbosity level? normal?
[15:12] <vsalento> error like always before
[15:13] <thelema> vsalento: what are the messages about?
[15:13] <vsalento> sliding buckets problems: guessed propability/interpolated crazy value
[15:14] <vsalento> newest version (from anon cvs -> might be the problem) - version 60171
[15:14] <toad_> <robilad> i have no idea how gcj-ed pure java big math fares vs. gmp, though - irrelevant, we have GMP on most platforms, right?
[15:14] <robilad> yep
[15:14] <toad_> <vsalento> is it on the todo list to use locally compiled libgmp if there is one? I need to replace to libjbigi-linux-none.so with one that uses shared lib since my arch is not supported. - what's your architecture vsalento?
[15:15] <hirvox> bbl, sleep
[15:16] <toad_> <vsalento> VIA Eden... not sure what it actually is but some i586 with some extras
[15:16] <toad_> <Iakin> vsalento: It will fallback..
[15:16] <toad_> <Iakin> To java code..
[15:16] <toad_> <Iakin> (since it will fail loading the .so that requires the other .so)
[15:16] <toad_> surely it'll use the generic x86 code?
[15:16] <toad_> <vsalento> I don't know... I just ran 'emerge gmp' and then when dynamically linked libjbigi to that the speed was 5 times better than using libjbigi-linux-none.so
[15:16] <toad_> woah
[15:16] <toad_> maybe there was one then
[15:16] <toad_> Iakin: did you do the targets?
[15:17] <toad_> if not, who did?
[15:17] <toad_> any chance of adding something suitable for such chips?
[15:17] <thelema> for the via eden, we need to check for the kernel support for the hardware RNG
[15:17] <thelema> :)
[15:18] <toad_> hehe
[15:18] <thelema> and hardware crypto
[15:18] <toad_> IIRC, the hardware RNG needs to be super-hashed
[15:18] <vsalento> propably... I'm not sure but there are 3dnow instrcutions also on this cpu that I don't think are available on normal i586 (I might be wrong)
[15:18] <toad_> i.e. it's not a very good RNG
[15:19] <toad_> can 3dnow be used by GMP? i thought it was all single precision floating point?
[15:19] <thelema> toad_: what I remember was just the opposite; that it's an excellent one.
[15:19] <KenMan> <Iakin> 2. Look in the classpath for the lib choosen above - i detected this, and was impressed by the logic :)
[15:19] <toad_> thelema: I was under the impression that 3dnow was strictly FP
[15:20] <toad_> thelema: ahhh
[15:20] <toad_> yeah
[15:20] <vsalento> it might have nothing to do with it ... it is possible that normal optimization for i586 would give same results
[15:20] <toad_> well I just saw something that was rather ambiguous
[15:20] <toad_> anyway that's kernel
[15:20] <toad_> i.e. we can't get to it on windows, probably, without considerable effort
[15:20] <wind789> the data is really flowing on unstable
[15:20] <toad_> and on linux, it's a separate device
[15:20] <vsalento> is there any reason to run 60171 ... it just fills log and cannot get connections?
[15:21] <wind789> anybody wanna downbload a 300kb test file I inserted?
[15:21] <thelema> true, it does need a non-sandard kernel patch to activate.
[15:21] <thelema> wind789: just post the URI and which network it's in.
[15:21] <wind789> unstable CHK@bL35x8zPm6Arup2hbeMoD1sRjygTAwI,uaQeOnu1lA0SQV4hROhnpg/Richard_Stallman-Free_Software.mp3
[15:22] * jay (jcl@ool-18bf6dac.dyn.optonline.net) Quit (".")
[15:24] <verl> wind789: dnf@20 htl
[15:24] <wind789> oh my
[15:25] <wind789> it collided twice..
[15:25] <wind789> keep trying, I'm reinserting
[15:27] * KenMan is finally collecting outgoingRequestInterval/min data... graphs to follow in a few hours...
[15:27] <wind789> I think the network is eating my node, 108 active transfers
[15:27] <thelema> wind789: bandwidth?
[15:27] <wind789> 41 kb/sec
[15:27] <KenMan> i think your node is gorging on the network ! so there !
[15:28] <wind789> actualy my bandwidth is averaging 61 kb/sec
[15:28] <wind789> 41 was instatanious
[15:28] * KenMan wonders when we lost AnnaMoo ...
[15:28] <thelema> .5 k/s per xfer. that's still a low lower than it should be...
[15:29] <KenMan> but better than it used to be !! think positive
[15:29] <wind789> 90 transmitting
[15:29] <thelema> :)
[15:29] <wind789> 108 was in and out
[15:29] <KenMan> see, it gets even better :)
[15:29] <vsalento> when was the 60172 committed to the CVS? I stopped the unstable node since the log file gets too big (errors, 1MB/min) but I cannot get any updates in anon-cvs
[15:29] <thelema> oh, a whole 2/3 K/s. That's great! </sarcasm>
[15:29] <wind789> lol
[15:30] <KenMan> hey, some manage to eek out a modest 50KB/s + ... right ?
[15:30] <wind789> think I should run a node with no keys? the node I'm currently running is loaded with stuff in the DS
[15:30] <toad_> <vsalento> is there any reason to run 60171 ... it just fills log and cannot get connections? - so run 60172 :)
[15:31] <vsalento> but I cannot get it from CVS!
[15:31] * thelema wants to know if most of those xfers are idle
[15:31] * KenMan wonders the same thing
[15:31] <wind789> anyway to find out?
[15:32] * toad_ fetches the RMS Song above referenced
[15:32] <KenMan> jab the toad in his belly, perhaps, to fix the mal-reported # xfers
[15:32] <toad_> hmm surprisingly small
[15:32] <wind789> indeed
[15:32] <wind789> the remix is 4 megs though hehe
[15:32] <toad_> KenMan: bugs will be dealt with in approximate priority order
[15:32] <KenMan> oh shit, just throw logic in my face, why don't you ;)
[15:33] <wind789> I'll insert the remix later
[15:33] <toad_> okay, starting up merged stable node...
[15:33] <thelema> KenMan: where are you getting the # xfers value?
[15:33] <toad_> maybe it'll work a bit better?
[15:33] <toad_> at least in terms of making connections?
[15:33] <KenMan> from OCM connections mode page
[15:34] <thelema> right from the html?
[15:34] <KenMan> yeah
[15:35] <toad_> okay, it loaded...
[15:35] * toad_ attempts to test pre-5087...
[15:36] * thelema starts looking into it
[15:36] <toad_> thelema: hmm?
[15:36] <toad_> ah
[15:36] <toad_> thanks
[15:38] <wind789> 100 megs transfered in 37 min on unstable
[15:38] <toad_> is that good?
[15:38] <toad_> are you using 60172?
[15:39] <wind789> well, it's maxing out my interconttion, so I guess it's good, yes 60172
[15:39] <verl> i think 60172 us
[15:39] <verl> *is behaving very well
[15:40] <verl> but have nothing to back this up with execpt the 'feel' of it :)
[15:40] <toad_> well, I'm trying pre-5087
[15:40] <wind789> was anybody able to download the song I inserted?
[15:40] <toad_> wind789: yes
[15:40] <wind789> cool
[15:40] <toad_> i remarked on how small it is
[15:40] * lostlogic_ (~lostlogic@node-402418b2.mdw.onnet.us.uu.net) has joined #freenet
[15:40] <wind789> thanks
[15:40] <wind789> true
[15:41] <verl> wind789: me too, a fine voice that man has
[15:41] <wind789> lol
[15:42] <wind789> in that case I'
[15:42] <toad_> opening connections... very slowly...
[15:42] <lolo-laptop> 527M in 1:24 on unstable code + stable network
[15:42] <wind789> *ll insert the 4 meg remix
[15:43] <toad_> hmmm
[15:44] <thelema> 300MB/hr -> 5MB/min
[15:44] <thelema> lolo-laptop: 10K/s bandwidth limit?
[15:44] <toad_> bbiab
[15:44] * KenMan apologizes for the 'poke the toad in his belly' remark
[15:45] <lolo-laptop> thelema: 50 or 60 k bw limit I think
[15:45] <thelema> lolo-laptop: really... is the 527MB just trailers?
[15:45] <lolo-laptop> thelema: mostly, I'd guess
[15:46] <KenMan> thelema - you can cut it down with sentData or receivedData if you must !!!
[15:46] <thelema> there are different diagnostics for all data xferred and just trailer data. i'm guessing that ...
[15:46] <thelema> KenMan: what?
[15:46] <KenMan> the percent of failed transfers
[15:47] <KenMan> so 527 would become 400 useful, maybe...
[15:47] <thelema> yeah, it's exactly this kind of thing that needs to be fixed before we bother doing anything with routing.
[15:48] <KenMan> we can't prevent people from restarting their nodes, it would be unamerican. Or un-something.
[15:49] <KenMan> Although i really doubt that can account for the bulk of it... not sure.
[15:49] <thelema> if restarted nodes are the bulk of xfer failures, then xfers are going on way way too long.
[15:49] <KenMan> somebody had a clever cron job that restarted fred every 30 minutes to 'maximize my success' :p
[15:56] * lostlogic (~lostlogic@node-402418b2.mdw.onnet.us.uu.net) Quit (Read error: 110 (Connection timed out))
[15:57] <toad_> compare the different sentData's
[15:57] <thelema> might as well have just killed all the ongoing transfers.
[15:57] <toad_> I don't think it's that bad now, most of the failures are to do with inserts
[15:58] * toad_ notes that fixing the size of files would reduce the length of time transfers are in flight for, and therefore reduce chances of such problems... and it would make new nodes, or nodes that aren't up for very long, more useful, because they'd get more routing experience faster
[15:59] <KenMan> just set 256 as max , that's the other option, right ??
[15:59] <lolo-laptop> toad_: you looked at that NetworkWritingThread thing or is it remaining low prio. for you ATM?
[15:59] <thelema> toad_: a protocol that doesn't have a 34K overhead for successful xfers would also do that.
[15:59] <toad_> thelema: it would? really?
[16:00] <thelema> in fact, a 34K overheadon successful xfers might be exactly what NGR is picking up to make it route poorly.
[16:00] <toad_> with a 300-500kB average file size?
[16:00] <toad_> StoreData's size isn't THAT big a deal
[16:00] <toad_> although I plan to deal with it
[16:00] * thelema thinks about it a bit more.
[16:00] <toad_> lolo-laptop: the what thing?
[16:00] <toad_> KenMan: yeah, but not as good
[16:00] <thelema> storedata's size is an amazing amount of overhead.
[16:01] <lolo-laptop> Jul 27, 2004 3:18:42 PM (freenet.transport.WriteSelectorLoop, Network writing thread, NORMAL): fixKeys added sun.nio.ch.SelectionKeyImpl@151716e
[16:01] <toad_> yeah, it's ~ 10%
[16:01] <toad_> that's not very nice
[16:01] <toad_> but it's not overwhelming
[16:01] * lolo-laptop gets lots of that
[16:01] <thelema> even if the average key is 400K, when you're surfing freesites, it is overwhelming.
[16:01] <toad_> lolo-laptop: hrrm
[16:01] <toad_> Iakin: any ideas on lolo's error?
[16:01] <toad_> thelema: hrrm
[16:01] <toad_> good point
[16:01] <KenMan> based on the keysize analysis (on both stable and unstable) the majority of activity is large items (1M pieces) moving around
[16:02] <thelema> 34 K of overhead on every mapfile, n every DBR
[16:02] <thelema> on every html file (tiny)
[16:02] <toad_> it doesn't wait for the storedata before showing the file... but it DOES still have to download it
[16:02] <thelema> on every tiny graphic and logo
[16:02] <toad_> but right now I have to merge...
[16:03] <toad_> hmmm
[16:03] <toad_> perhaps my stable node is wedged?
[16:03] <KenMan> mine (5086) is finally up to what i could call 'smooth sailing'
[16:03] <toad_> yup, definitely stuck
[16:03] <toad_> not responding
[16:04] <toad_> how come this only happens on stable despite similar codebase?
[16:04] <KenMan> in fact, the only thing I would complain about with 5086 at this point, is the wiggliness of queries and MRIs... and maybe too few peers.
[16:04] <toad_> hmm, cpu usage is maxed out...
[16:04] <toad_> and it started a little while ago...
[16:05] <toad_> probably after i started.....hmmm
[16:05] * toad_ turns off unstable node
[16:07] <toad_> hmmm
[16:07] <toad_> it doesn't seem to be a single thread in a loop
[16:07] <toad_> except for the VM thread, which we can do nothing about
[16:08] <toad_> the VM thread is the dominant thing in top... always...
[16:09] <toad_> GCs every 2 seconds or so, but they're all very short - 0.1 seconds at most
[16:09] <toad_> so that's not it
[16:09] <toad_> actually noo, it isn't gcing
[16:10] <toad_> or is it?
[16:10] <toad_> no, no GCing
[16:10] <toad_> ok
[16:10] <toad_> maybe it's constantly retrying connections/
[16:10] <toad_> ?
[16:11] <KenMan> you are making stack dumps ? and no help ?
[16:11] <lolo-laptop> :-(, all my routes are backed off due to 3 30 thread splitfile attempts...
[16:11] <toad_> KenMan: no help
[16:11] <toad_> hmmm
[16:11] <toad_> eventually it responded
[16:12] <toad_> but it took aaaaaages
[16:12] <lolo-laptop> Attempts were made to contact 0 nodes. 59 backed off.
[16:12] <KenMan> check the number of connections +/- diagnostic
[16:12] * KenMan is relegated to 'debugging' via the diags :)
[16:14] <toad_> very likely hardly any connections
[16:15] <toad_> hmmm
[16:15] <toad_> 15 conns
[16:15] <toad_> woah
[16:15] <toad_> Env says:
[16:15] <toad_> Checkpoint: Opening connection 48
[16:15] <toad_> doh
[16:15] <toad_> wrong page
[16:15] <thelema> toad_: this code is funny:
[16:15] <thelema> + int l = bitLength() + 1;
[16:15] <toad_> Checkpoint: Check for connections to open 1
[16:15] <toad_> Checkpoint: Opening connection 32
[16:15] <thelema> + int targetLength = l / 8 + ((l % 8 != 0) ? 1 : 0);
[16:15] <toad_> freenet.interfaces.LocalNIOInterface$ConnectionShell 3
[16:15] <toad_>
[16:15] <toad_> okay
[16:16] <toad_> well it's possible.. hmmm
[16:16] <toad_> KenMan: if you have no great insights, I'll restart with the "backoff if conn lasts less than 20 seconds" patch merged
[16:17] <thelema> toad_: targetLength = bitLength() / 8 + 1
[16:17] <toad_> ?
[16:17] <toad_> thelema: no
[16:17] <toad_> if bitLength % 8 == 0
[16:17] <toad_> errr
[16:17] <toad_> no if (bitLength+1) % 8 == 0 ?
[16:17] <toad_> then bitLength/8 is less 1.. hrrm
[16:17] <toad_> well change it if you like
[16:18] <thelema> :)
[16:18] <toad_> okay
[16:18] <toad_> 7 connections open
[16:18] <toad_> 32 openers running
[16:18] <toad_> uptime must be an hour or so now
[16:18] <thelema> 32 openers running after an hour?
[16:19] <toad_> suggests backoff isn't working...
[16:19] <toad_> sshall i restart with the backoff-if-conn-lasts-less-than-20-seconds change?
[16:19] <toad_> it'll take a little while for it to get messed up again...
[16:19] <toad_> if indeed it does...
[16:20] <toad_> can i observe this unique state in any other way before restarting?
[16:20] <lolo-laptop> 8 openers, 60 connectinos open after 2 hours...
[16:21] <toad_> hmmm
[16:21] <toad_> unknown number of openers
[16:21] <toad_> ah
[16:21] <toad_> 12 connections open
[16:22] <toad_> 12 transfers active (9 up, 3 down)
[16:22] <toad_> uptime 48 minutes
[16:22] <toad_> still waiting for the Env page
[16:22] <toad_> memory in use: ~ 107MB
[16:23] <toad_> 24 connection openers
[16:23] <toad_> 43 threads in DataRequest
[16:23] <toad_> eeek
[16:24] <toad_> hmmm
[16:24] <toad_> lots of ConnectionOpeners waiting for ConnectionOpenerManager
[16:24] * KenMan has no great insights :(
[16:24] <toad_> lots of DataRequests waiting for log...
[16:24] <wind789> unstable CHK@FwDwUWZKHDd3IeAzR95gCSifKVwMAwI,URWySMGDJK2BUbFK4s1JkQ/Richard_Stallman-Free_Software(remix).mp3
[16:24] <wind789> 4.19 megs
[16:24] <KenMan> toad_ 5086 incorporates "communicate mRI as often as possible" , right ?
[16:24] * oierw pines for a testnet where you can se why the request is blocked
[16:24] <toad_> COM is locked...
[16:25] <toad_> apparently waiting for logger
[16:25] <toad_> what a surprise ;)
[16:25] <verl> a 22mb log of these: 2004-jul-27 22:41:33 (freenet.transport.WriteSelectorLoop, Network writing thread, NORMAL): fixKeys added sun.nio.ch.SelectionKeyImpl@1962533
[16:25] <toad_> possibly just contention...
[16:26] <thelema> toad_: fixKeys is logging a lot
[16:26] <toad_> and the logger thread is in DateFormat
[16:27] * toad_ doesn't see enough high-level logging to cause the problem...
[16:28] <toad_> more likely to be low-level logging anyway...
[16:30] * toad_ unsets logLevelDetail
[16:30] <toad_> and sets logLevel up to NORMAL
[16:31] <toad_> and merges the backoff if conn up for less than 20 secs code
[16:32] * toad_ starts node again...
[16:32] <KenMan> toad_: 5086 incorporates "communicate mRI as often as possible" , right ?
[16:32] <toad_> KenMan: not quite
[16:33] <toad_> but pretty often
[16:33] <KenMan> more than it used to. Okay.
[16:33] <toad_> next build will include it on every packet
[16:33] <toad_> 5086 includes it on every FNP message, as do builds going back a looong way
[16:33] <KenMan> oh, thank you.
[16:33] <toad_> 5087 only does this if talking to 5084 or earlier nodes
[16:35] <toad_> woah, I have 13 connections ALREADY!
[16:35] <toad_> aww, just lost one
[16:35] <toad_> down to 12...
[16:37] <vsalento> I hate this anon-cvs lag... what does it require to get non-lagging cvs access?
[16:37] <toad_> hmmm
[16:37] <toad_> strange RNFs...
[16:38] <toad_> Couldn't retrieve key: SSK@rBjVda8pC-Kq04jUurIAb8IzAGcPAgM/TFE//
[16:38] <toad_> Hops To Live: 15
[16:38] <toad_> Error: Route Not Found
[16:38] <toad_> Attempts were made to contact 0 nodes.
[16:38] <toad_> * 0 were totally unreachable.
[16:38] <toad_> * 0 restarted.
[16:38] <toad_> * 0 cleanly rejected.
[16:38] <toad_> * 10 backed off.
[16:38] <toad_> i'm sure i had 14 open conns at the time...
[16:38] <toad_> Error: Route Not Found
[16:38] <toad_> Attempts were made to contact 0 nodes.
[16:38] <toad_> * 0 were totally unreachable.
[16:38] <toad_> * 0 restarted.
[16:38] <toad_> * 0 cleanly rejected.
[16:38] <toad_> * 1 backed off.
[16:38] <toad_> WTF?!?!
[16:39] <wind789> stable?
[16:39] * Ash-Fox (Hal-9000@aaq112.neoplus.adsl.tpnet.pl) Quit (Remote closed the connection)
[16:39] <lolo-laptop> * 2 cleanly rejected.
[16:39] <lolo-laptop> * 54 backed off.
[16:39] <lolo-laptop> blah.
[16:39] * greycat (~wooledg@192.35.79.70) Quit ("Lick Bush in '04")
[16:42] <wind789> what version are you guys running?
[16:44] * lolo-laptop running latest unstable CVS on stable network.
[16:44] <toad_> hmmm
[16:44] <toad_> my outgoing MRI is rapidly rising...
[16:44] <KenMan> from ? to ?
[16:44] * Ash-Fox (Hal-9000@aaq112.neoplus.adsl.tpnet.pl) has joined #FreeNET
[16:44] * bluephile (bluephile@69-160-205-193.clvdoh.adelphia.net) Quit (Client Quit)
[16:45] <toad_> 7/27/04 9:52:00 PM BST 332.97631038826694 239.64808622934123 -1.0 1243.0564545208 902
[16:45] <toad_> 7/27/04 9:53:00 PM BST 489.9078948967802 424.1918900397577 -1.0 3224.887839726312 2593
[16:45] <toad_> 7/27/04 9:54:00 PM BST 1152.2023563347113 895.680070117491 -1.0 3634.2569808468793 2507
[16:45] <toad_> 7/27/04 9:55:00 PM BST 2085.7416219036872 1724.9746399474886 -1.0 5147.914738009691 2081
[16:45] <toad_> 7/27/04 9:56:00 PM BST 2971.8009677376845 2386.1786501686674 -1.0 6629.625679608179 1972
[16:45] <toad_> 7/27/04 9:57:00 PM BST 3908.575175213232 2763.935503641307 -1.0 7163.475809244128 1713
[16:45] <toad_> 7/27/04 9:58:00 PM BST 4046.5140419832287 2909.077733388863 -1.0 7719.837170184368 1528
[16:45] <toad_> 7/27/04 9:59:00 PM BST 16484.124341531973 82567.48949457896 -1.0 600000.0 1787
[16:45] <toad_> 7/27/04 10:00:00 PM BST 15812.891648886553 77076.52122395531 -1.0 600000.0 1876
[16:45] <thelema> toad_: probably because it has nowhere to send requests to
[16:45] <toad_> 7/27/04 10:01:00 PM BST 16416.5098678191 78362.68924722483 -1.0 600000.0 1529
[16:45] <toad_> 7/27/04 10:02:00 PM BST 16879.914441067438 77814.52745039274 -1.0 600000.0 1606
[16:45] <toad_> seems stable on ~ 16 seconds for now...
[16:45] <toad_> thelema: huh?
[16:45] <KenMan> oh, you need to stop looking at it for about an hour...
[16:45] <thelema> 1 backed off -> bad mojo
[16:45] <toad_> i wonder why the sudden jump
[16:46] <toad_> thelema: it doesn
[16:46] <toad_> it doesn't take that into account in load calculation
[16:46] <KenMan> my outRI is 'randomly' drifting from 100s to 600s
[16:46] <KenMan> mRIs take a long time to settle in, if they ever do...
[16:46] <thelema> it doesn't directly, but it's got to indirectly.
[16:47] * thelema tries to think of how...
[16:49] <toad_> okay, i think that the problems were just because of logging...
[16:50] * toad_ commits 5067
[16:51] <KenMan> just make sure that Version.java says "5087"
[16:51] <KenMan> or are we trying to cover up our mistakes ;) ?
[16:51] * bluephile (bluephile@69-160-205-193.clvdoh.adelphia.net) has joined #freenet
[17:01] * Superfan is now known as TLF
[17:03] * Hory (~Miranda@81.196.25.110) Quit ("Miranda IM! Smaller, Faster, Easier. http://miranda-im.org")
[17:08] <toad_> hehe
[17:08] * toad_ changes topic to 'Freenet: Taming the World's Largest Tamagotchi | Upgrade to 5087 | Unstable: Upgrade to 60172 | Logs: http://newton.matcmp.ncc.edu/~lockej/freenet/chanlog/'
[17:09] <toad_> 5087 is out!
[17:09] <toad_> everyone upgrade!
[17:12] <Iakin> [21:32] <toad_> Iakin: did you do the targets?
[17:12] <Iakin> [21:33] <toad_> if not, who did?
[17:12] <Iakin> [21:33] <toad_> any chance of adding something suitable for such chips?
[17:12] <Iakin> Yes, I consider doing it :)
[17:12] <toad_> Iakin: I'm writing a mail on devl
[17:12] <toad_> outlining which targets i'd like
[17:13] <Iakin> [21:35] <toad_> can 3dnow be used by GMP? i thought it was all single precision floating point?
[17:13] <Iakin> [21:35] <thelema> toad_: what I remember was just the opposite; that it's an excellent one.
[17:14] <toad_> Iakin: that was in response to another thing
[17:14] <Iakin> I think that SSE (on the Athlon XP or better) actually is faster than 3dnow on it..
[17:14] <Iakin> I read that somewhere
[17:14] <toad_> SSE probably is
[17:14] <toad_> SSE2 certainly is
[17:14] <toad_> that's why AMD implemented SSE2 (and enhanced it too) in the K8 :)
[17:15] <thelema> Iakin: that was in response to his comment about the C5 Ezra processor's RNG
[17:17] * Hory (~Miranda@81.196.25.110) has joined #FreeNet
[17:18] <lolo-laptop> JVM CRASH, YAY!!
[17:18] <Iakin> [22:17] <toad_> Iakin: any ideas on lolo's error?
[17:19] <Iakin> Well.. not staightaway.. possibly that the key became ready after the selection?
[17:19] <lolo-laptop> sigh, and my sheet isn't using native FEC for some reason...
[17:19] <Iakin> Or that fixKeys checks for the wrong op?
[17:19] <Iakin> or something
[17:19] <lolo-laptop> *DLs latest freenet-ext incase I messed something up*
[17:20] <Iakin> Would the key appear during the next select if it wasn't fixKey'ed?
[17:23] * TLF (francisco@29.Red-81-40-113.pooles.rima-tde.net) Quit ("http://www.vitaliano.esp.cc || Adi?s a todos y a todas y que les vaya bien.")
[17:24] * salahx (salahx@sc1-24.217.174.147.charter-stl.com) has joined #freenet
[17:29] <toad_> hmmm
[17:29] <toad_> my outgoingRequestInterval on the stable node is VERY high...
[17:29] <toad_> ~ 400,000ms...
[17:30] <thelema> lots of xfers?
[17:30] <thelema> routing succeeding -> higher success rate -> higher load -> higher mRI -> higher routedToChoiceRank -> worse routing
[17:31] <toad_> eh?
[17:31] <toad_> higher mRI on my node leads to worse routing how?
[17:33] <toad_> globally maybe
[17:33] <toad_> but locally?
[17:33] * Ash-Fox (Hal-9000@aaq112.neoplus.adsl.tpnet.pl) Quit (Client Quit)
[17:33] <toad_> thelema: what's your opinion on the KenManHack?
[17:34] <toad_> bbl
[17:34] <KenMan> so you didn't have to wait an hour for MRI to go high ??
[17:34] <thelema> the higher mRI -> higher rTCR is a intra-node phenomenon; your node having a high mRI increases other nodes' rTCR
[17:34] <thelema> KenMan: what's the KenManHack again?
[17:34] <KenMan> it is supposed to regulate MRI in response to success rates...
[17:35] <thelema> mri = mri * success %?
[17:35] <salahx> yep!
[17:35] <thelema> so it makes the connection (higher success rate -> higher load) even stronger?
[17:35] <KenMan> it reserves enough bandwidth that if 100% of queries succeeded, we would not be underwater. By reducing total query volume, it should reduce rTCR and help boost success.
[17:36] <thelema> I approve of much faster transfers.
[17:36] <KenMan> well, actually, instead of 100%, I think toad used a different target. Like best success ever seen or something like that.
[17:36] <thelema> I don't think reducing total query volume works like you think.
[17:36] <thelema> s/works/will work/
[17:36] <KenMan> i don't either, but it was just an idea I tossed out there.
[17:37] <thelema> toad_: my opinion on teh KenManHack is that I'm against it.
[17:37] <KenMan> hehe
[17:38] <KenMan> the theory is sound (it is so basic), but whether it can have actually affect success rates is dubious.
[17:38] <thelema> We *need* to increase fred's load handling capacity. That's my only hope for dealing with the constant overload.
[17:38] <KenMan> the premise is that lower query rates produces better routing. That is not proven by any means...
[17:39] * Hory (~Miranda@81.196.25.110) Quit (Read error: 60 (Operation timed out))
[17:40] <KenMan> although this picture kind of supports it... http://mywebpages.comcast.net/jkcorson/FreeNet/5086c.png
[17:40] <KenMan> blue sorta causes green, which mostly causes brown, which determines orange.
[17:41] <KenMan> brown can make orange go high faster than it can make it go low. That's my *opinion* anyway
[17:41] <thelema> ther's no green
[17:41] <KenMan> scroll to the right
[17:42] <thelema> again, there's no green
[17:42] * m[a]zzanet (~mazzanet@mazzanet.registered) has joined #freenet
[17:42] <KenMan> how many lines can you make out ?
[17:42] <thelema> 4
[17:42] <KenMan> there should be 6 if you scroll to the rightmost hand side
[17:43] * m[a]zzanet is now known as mazzanet_
[17:43] <KenMan> try a forced reload, perhaps your browser cache is screwed. How many do you see in the upper-left legend ?
[17:43] <thelema> red, blue, orange, brown. there's only those in the legend
[17:43] * mazzanet (~mazzanet@mazzanet.registered) Quit ("Read error: Connection reset by Telstra")
[17:43] <KenMan> reload while holding down the shift key
[17:43] <thelema> ah, now I see the last two.
[17:44] <thelema> I had forgot about browser cache
[17:44] * mazzanet_ (~mazzanet@203-173-44-61.dyn.iinet.net.au) Quit (Client Quit)
[17:45] * mazzanet (~mazzanet@203-173-44-61.dyn.iinet.net.au) has joined #freenet
[17:45] <KenMan> anyway, the best observation is the brown-orange relationship. At least that appears to be working / not completely broken.
[17:46] <KenMan> i just wish the brown line was more level, holding around 20 qpm ...
[17:46] <KenMan> or whatever the smoothed average would be !
[17:47] <thelema> Think of the granularity of the sampling interval
[17:47] <thelema> and the granularity of requests.
[17:48] <thelema> if requests take up too much resources, it's impossible to make the level smooth on a fine interval
[17:48] <KenMan> yes, it would only spike even more if I (could) sampled at a finer resolution
[17:48] <KenMan> i just want "smoother" ... :)
[17:49] <thelema> Just like packing a barrel with bowling balls. Versus filing it with sand.
[17:49] <KenMan> then, it think it is possible to use smaller bowling balls...
[17:50] <KenMan> are you making the analogy of bowling balls to key size, or just individual requests ?? I assume you mean requests
[17:51] <thelema> just the weight of requests. Smaller key sizes are smaller units, but also if we didn't work as hard to fulfill requests, that'd compare to smaller bowling balls.
[17:51] <thelema> I also wonder if it'd be possible to make smaller requests flow easier...
[17:52] <thelema> mRI treats all requests the same.
[17:52] <thelema> what if instead of mRI there was a system that allowed 8 32K requests _or_ 1 1MB request
[17:52] <thelema> not quite equal because there's overhead on the short requests
[17:53] <KenMan> yes, you ARE considering keysize. Ok. I like your thinking... and we can see the keysize in the URI.
[17:53] <KenMan> but success rates come into play also , especially when they are this low.
[17:54] <thelema> not in the URI, but in the routing key
[17:54] <KenMan> aren't they the same thing ?
[17:54] <thelema> true; if we had to, we could charge exactly based on an estimate of expected work. more for higher HTL, etc.
[17:55] <thelema> they aren't the same thing; URI is a CHK@blah or SSK@blah/blah or KSK@blah
[17:55] <thelema> routing key is h(h(chk@blah))
[17:55] <thelema> or something equivalent for SVKs
[17:55] <thelema> the routing key is a 23-bit value.
[17:55] <thelema> *byte
[17:55] <KenMan> shit, we aren't really doing that, i didn't think. It was discussed, but not implemented.
[17:56] <thelema> no, that's *another* level of hashing that's been proposed.
[17:56] <KenMan> CHK@hash,key is the current system, right ?
[17:56] <thelema> I did get it wrong for CHK
[17:57] <KenMan> phew, I was going to permanently walk away if I had that bad of an understanding...
[17:57] <thelema> CHK@blah,key -> routing key is h(blah)
[17:57] <thelema> no, maybe I did get it right...
[17:57] <KenMan> that is not implemented today.
[17:57] <thelema> the encryption key can't be derived from the routing key for CHKs
[17:57] <KenMan> we talked about the hash of the hash of the cash of the hash... but i thought all we did was talk.
[17:57] <thelema> and the actual ClientKey can't be derived from either.
[17:58] <thelema> given <DATA>, the encryption key is H(data)
[17:58] <KenMan> you are on SSK level. I am still at CHK level.
[17:58] <KenMan> wha ?? ?
[17:59] <KenMan> I thought the key was randomly chosen. It really should be anyway
[17:59] <KenMan> I am very worried about one of us, just not sure which one yet.
[17:59] <thelema> the routing key is H(H(data))
[17:59] <KenMan> stop filling my head with your lies !!! ;)
[17:59] <thelema> the key can't be randomly chosen
[18:00] <KenMan> not routing key, crypto key
[18:00] <thelema> if it is, then CHK(BritnetSpears1.mp3) != CHK(BritnetSpears1.mp3)
[18:00] <thelema> again, crypto key can't be either.
[18:00] <KenMan> the ,decrypt key portion can be random, but the CHK@hash part can't be
[18:00] <thelema> if the crypto key is randomly chosen, then the encrypted data will be different
[18:01] <thelema> and two insertions of the same file will be encrypted differently and can't be verified
[18:01] <KenMan> I see your point and why it might be that way. Yes. I feel very disconnected from the basics...
[18:02] <thelema> Read up on keys from freenethelp; I wish I had time to finish explaining this to you.
[18:02] <thelema> bbl
[18:02] * thelema is now known as thelema|aikido
[18:02] <KenMan> Wait, two insertions of the same file would just have different decrypt keys, not different hashes. Anyway, you is probably the more knowledgeable one here :)
[18:03] * KenMan figures he should just walk through the entire process in the code... some day when he has an entire day to kill.
[18:04] <KenMan> oh : the key never floats around the network, so it would be unfortunate to have 2 insertions with different keys using the same CHK@hash .
[18:04] <KenMan> it all sounds vaguely familiar. No one ever works on this part of freenet because it just works.
[18:05] <KenMan> wow, i really am off base in how I think a lot of things work :(
[18:13] <KenMan> I swear I thought this post was pure sarcasm - http://article.gmane.org/gmane.network.freenet.devel/6947
[18:14] <KenMan> "it is stored in the hash of the hash of the hash," <- people were throwing around a lot of ideas at the time. I thought this person was being sarcastic of the whole discussion, but he WAS serious.
[18:14] <salahx> yeah the CHK stuff is a nightmare
[18:15] <KenMan> I figured that out later as people made serious replies to him...
[18:15] <KenMan> well, it is going to take some doing to replace the conceptions I've been carrying around for a year now, that is all...
[18:18] <iip_i2p> <Newsbyte> toad, you there? I tried out the new build
[18:21] * KenMan doesn't hear any bitching after less than an hour of experience might have been gathered ... ;)
[18:21] <KenMan> so what do you think Newsbyte ?
[18:25] * Ash-Fox (Hal-9000@aae32.neoplus.adsl.tpnet.pl) has joined #FreeNET
[18:33] * toad_ would prefer to fix key size rather than mess around with weighing requests by size etc
[18:34] <toad_> thelema|aikido: why are you against KenManHack?
[18:35] <toad_> <thelema> the key can't be randomly chosen - it can if you don't have to fetch it
[18:35] <iip_i2p> <Newsbyte> well, I think it's basically the same
[18:35] <toad_> e.g. if you are trying to do a DNF flood
[18:35] <iip_i2p> <Newsbyte> crappy, thus
[18:36] <iip_i2p> <Newsbyte> I started with 20% and fell to 1-2% again
[18:36] <toad_> 20% what?
[18:37] <toad_> that sounds like a bogus effect to me... requestSuccessRatio starts high because there are many local requests and few remote ones
[18:37] <iip_i2p> <Newsbyte> though I do get activelinks this time (not that they go to anything)
[18:37] <toad_> and goes low because local requests decrease and remote requests increase
[18:37] <iip_i2p> <Newsbyte> statusbar
[18:37] <toad_> what statusbar?
[18:38] <iip_i2p> <Newsbyte> the statusbar on the webinterface
[18:38] <toad_> umm, why are you concerned about that?
[18:38] <toad_> how many connections do you have open?
[18:39] <toad_> mine says 39%
[18:39] <toad_> but that's just load
[18:39] <toad_> yes ideally it would be 100%
[18:39] <iip_i2p> <Newsbyte> o, istelf, it wouldn't, but I've noticed the pattern sine 5085, and when that statusbar remains that puny, it's not biding well
[18:39] <toad_> but i don't see "my node's load is 2%" as a big concern...
[18:40] <toad_> well how many connections do you have?
[18:40] <toad_> what's your requestSuccessRatio?
[18:40] <toad_> what's your routingSuccessRatio?
[18:40] <iip_i2p> <Newsbyte> Connections open (Inbound/Outbound/Limit) 7 (0/7/200)
[18:40] <toad_> how many queries is your node accepting per hour roughly?
[18:40] <toad_> Newsbyte: okay, that's bad
[18:40] <toad_> when did you restart the node?
[18:41] <toad_> and that's the CURRENT number?
[18:41] <toad_> not the number an hour ago?
[18:41] <iip_i2p> <Newsbyte> ermm...when 8087 came out? ;-)
[18:41] <iip_i2p> <Newsbyte> ermm...when 8087 came out? ;-)
[18:41] <iip_i2p> <Newsbyte> 5087
[18:41] <iip_i2p> <Newsbyte> 5087
[18:41] <toad_> when was that?
[18:42] <iip_i2p> <Newsbyte> ermm...when you announced it on devl? I dunno, some hours ago
[18:43] <toad_> uh, so how come my stable node has 61 connections?
[18:43] <toad_> and looks reasonable otherwise?
[18:44] <toad_> that suggests it's not an inherent problem in the build
[18:44] <toad_> is your node getting requests?
[18:44] <iip_i2p> <Newsbyte> no, it's NOT because it should get better connected. I've restarted many times a build, and it never went the way it goes now since 5085.
[18:44] <salahx> unstable is looking good here but its onyl been an hour
[18:44] <toad_> uh?
[18:44] <toad_> if your node has only connected to 7 other nodes, then it's bound to suck
[18:45] <toad_> that is inevitable
[18:45] <iip_i2p> <Newsbyte> that is possible, yet, it is on my node, and at least with 5085 i wasn't the only one
[18:45] <toad_> what does http://127.0.0.1:8888/servlet/nodestatus/diagnostics/localQueryTraffic/hour say?
[18:45] <toad_> has your node been sent any requests?
[18:45] <iip_i2p> <Newsbyte> then again, you don't use windows, do you, toad?
[18:45] <toad_> many requests?
[18:45] <toad_> not on this computer
[18:46] <iip_i2p> <Newsbyte> ermm, you didn't see that, I suppose? damn '!say'
[18:46] <toad_> my stable node has 60 connections and is serving 10qph at a requestSuccessRatio of ~ 5%
[18:46] <toad_> and routingSuccessRatio of ~ 1.5%
[18:47] <toad_> that's a bit low but i expect it to recover as the network recovers
[18:47] <toad_> err 10kqph
[18:47] <toad_> :)
[18:48] <iip_i2p> <Newsbyte> ratio is 1.0
[18:48] <toad_> Newsbyte: counts?
[18:49] <toad_> "tries" or "successes"
[18:49] <iip_i2p> <Newsbyte> 7/28/04 1:00:00 AM CEST 1878 1878 1.0
[18:49] <toad_> on seven connections?!?!
[18:49] <toad_> woah
[18:49] <iip_i2p> <Newsbyte> yes, yes, I already pasted everything, but do to eyekon, you couldn't see it
[18:49] <toad_> they must be seriously desperate...
[18:50] <toad_> can you please re-check http://127.0.0.1:8888/servlet/nodeinfo/networking/ocm?setMode=Connection ?
[18:50] <iip_i2p> <Newsbyte> Connections open (Inbound/Outbound/Limit) 7 (0/7/200)
[18:51] <toad_> obviously they are... my node is returning an outgoingRequestInterval averaging at 200,000...
[18:51] <iip_i2p> <Newsbyte> Transfers active (Transmitting/Receiving) 2 (2/0)
[18:51] <iip_i2p> <Newsbyte> Data waiting to be transferred 413 Bytes
[18:51] <iip_i2p> <Newsbyte> Total amount of data transferred 5,549 KiB
[18:51] <toad_> Newsbyte: I'm amazed that the network sends you 2kqph when you only have 7 connections
[18:52] <toad_> have you reseeded?
[18:52] <iip_i2p> <Newsbyte> The fact I only get 7 connections is also the reason it's crappy
[18:52] <iip_i2p> <Newsbyte> how many f- times do I have to reseed?
[18:53] <toad_> I don't know
[18:53] <toad_> have you reseeded since 5087 came out?
[18:53] <toad_> lets see...
[18:53] <iip_i2p> <Newsbyte> I reseede with 5084, twice with 5085, did an auto-update for 5087, and now I have to reseed again?
[18:54] <toad_> well the reason your node sucks is that it has no connections
[18:54] <toad_> your alternative is to just leave it running
[18:54] <toad_> maybe it will gather more connections
[18:54] <iip_i2p> <Newsbyte> that was the same reason it was crap with 5085
[18:54] <toad_> if it's processing 2kqph, it's an active part of the network
[18:54] <toad_> even if a minor one
[18:55] <toad_> well, i do think the main connection bugs are fixed with this build
[18:55] <iip_i2p> <Newsbyte> but while shortly improving after reseeding, it then fell back to nothing
[18:55] <toad_> but I know that it may be a problem e.g. with existing 5085/5086 nodes
[18:55] <iip_i2p> <Newsbyte> as it is/was now
[18:55] <toad_> well those are the obvious options
[18:56] <iip_i2p> <Newsbyte> and I don't understand anyway; shouldn't I get new nodes as seednodes from the 7 that do work?
[18:56] <iip_i2p> <Newsbyte> and I don't understand anyway; shouldn't I get new nodes as seednodes from the 7 that do work?
[18:56] <toad_> yes
[18:56] <toad_> but it may take a while
[18:56] <wind789> I had no idea unstable was such a busy network
[18:56] <toad_> especially if you have difficulty connecting to them
[18:56] * ysothoth (~ysothoth@68-174-22-46.si.rr.com) Quit ("using sirc version 2.211+KSIRC/1.3.10")
[18:57] <toad_> wind789: it's amazing, considering there's virtually no content
[18:57] <wind789> lol
[18:57] <toad_> wind789: what makes you say that anyway?
[18:57] <toad_> Newsbyte: and you will have difficulty connecting to them
[18:57] <toad_> Newsbyte: would you like me to email you some locally generated seednodes?
[18:58] <salahx> what the heck are these in my logs: fixKeys added sun.nio.ch.SelectionKeyImpl@cf07b3 ?
[18:58] <wind789> toad_: My node's bandwidth has been maxed for 4 hours, 640 megs transfered so far
[18:58] <iip_i2p> <Newsbyte> fine, I'll reseed, again. I'll tell you what will happen: the first few minutes, it will work considerably better, then I'll see the statusbar drop to 1%, and connections will be crap again. I've been over this twice with 5085, and it will be the same now.
[18:58] <toad_> salahx: either a logging bug or an NIO bug
[18:58] <toad_> I don't see what the statusbar has to do with anything
[18:59] <toad_> I'm not sure what you want me to do about it
[18:59] <iip_i2p> <Newsbyte> I found it to be a indicative sign
[18:59] <wind789> 18,000 requests
[18:59] <iip_i2p> <Newsbyte> ermm...to fix it? ;-)
[18:59] <toad_> well, your node IS processing requests...
[18:59] <toad_> Newsbyte: I can't fix it if I can't reproduce it
[19:00] <toad_> afaics your node isn't newbie, therefore it isn't announcing
[19:00] <iip_i2p> <Newsbyte> goodie! Now it just has to be able to retrieve even the puniest of flogs...
[19:00] <wind789> unique contacts: 126
[19:00] <toad_> it probably has a routing table full of 5085 and 5086 nodes that it can't connect to
[19:00] <toad_> because of their bugs
[19:00] <toad_> otoh, most of my RT is 5084
[19:00] <toad_> so your node maybe ought to be able to connect to 5084s
[19:00] <iip_i2p> <Newsbyte> with a testnetwork, you could
[19:01] <toad_> not necessarily
[19:01] <toad_> we tested this code on unstable for MONTHS
[19:01] <toad_> it's probable that the network hasn't adjusted to the 5085 routing changes yet
[19:02] <iip_i2p> <Newsbyte> unstable is not a real testnetwork in the strict sense
[19:02] <toad_> as far as your node is concerned... it sucks because it has only 7 connections (how big's your RT?). it's getting an amazingly high number of reqs considering that.
[19:02] <iip_i2p> <Newsbyte> anywayz, I'll reseed, but things will go as I've predicted, mark my words
[19:03] <toad_> presumably it tries to connect to nodes in its RT, and they fail, either because of overload caused by 5085/6 suckage, or because of 5085/6 suckage in connections itself
[19:03] <toad_> what does http://127.0.0.1:8888/servlet/nodestatus/nodestatus.html say?
[19:03] <iip_i2p> <Newsbyte> what's the url again?
[19:03] <toad_> i have:
[19:03] <toad_> Connection Attempts 3182
[19:03] <toad_> Successful Connections 66
[19:03] <toad_> what url?
[19:04] <iip_i2p> <Newsbyte> I just disconnected my node
[19:04] <iip_i2p> <Newsbyte> for the seednodes.ref
[19:04] <toad_> 1 in 50 connections succeeds... why? because of the 5085/6 suckage with connections, probably
[19:04] <toad_> http://freenetproject.org/snapshots/
[19:04] <wind789> ever consider a third branch for longer term beta testing of future stable relases?
[19:04] <toad_> I can give you local testing
[19:04] <toad_> wind789: it's called "stable" :)
[19:05] <KenMan> 5086 conns,backoff, MRIs ... -> http://mywebpages.comcast.net/jkcorson/FreeNet/5086c.png
[19:05] * KenMan restarts with 5087 now...
[19:05] <toad_> KenMan: how many conns?
[19:05] <wind789> yeah but it snot stable if it's broken
[19:05] <KenMan> it only has about 60 according to that graph
[19:06] <iip_i2p> <Newsbyte> I was thinking...couldn't my router/NAT thingy have anything to do with it? It worked fine with 5084, and the later builds were supposed to work *better* with routers, but... (?)
[19:06] <KenMan> the current graph runs past 23:00 so reload around your browser-cache if you don't see that on the bottom right
[19:07] <toad_> KenMan: that graph is a great argument for the KenManHack
[19:07] <KenMan> why doesn't the blue shoot upwards for several minutes after queries have totally overtaken backoff level ??
[19:08] <KenMan> anyway, let's see what 5087 does...
[19:09] * salahx gets out the fire extingisher
[19:09] <KenMan> backoff and query level rise and fall in close sync, but if queries fell to zero and stayed there, how long would backoff (orange) take to reach zero ??
[19:10] <KenMan> and why does query level go up and down quite so much ?
[19:10] <iip_i2p> <Newsbyte> great...now it even can't get started...
[19:12] <KenMan> can 5087 use jbiggy ?
[19:12] * toad_ (toad@82-32-16-91.cable.ubr03.azte.blueyonder.co.uk) Quit ("using sirc version 2.211+KSIRC/1.3.10")
[19:12] <KenMan> bye toad...
[19:13] * jay (jcl@ool-18bf6dac.dyn.optonline.net) has joined #freenet
[19:13] <iip_i2p> <Newsbyte> god damn, it just hangs there ad infintum
[19:13] * toad_ (toad@82-32-16-91.cable.ubr03.azte.blueyonder.co.uk) has joined #freenet
[19:13] <KenMan> hi toad, how are you today ?
[19:13] <toad_> sorry ppl, my fscking irc client crashed
[19:13] <toad_> did I miss anything interesting?
[19:14] <iip_i2p> <Newsbyte> that's strange, java didn't start, that's why, I suppose
[19:14] <toad_> eh?
[19:14] <iip_i2p> <Newsbyte> yes, you did
[19:14] <toad_> what did I miss?
[19:14] <iip_i2p> <Newsbyte> after reseeding, freenet didn't start up
[19:14] <toad_> uhh, you did close the node down first?
[19:15] <iip_i2p> <Newsbyte> the bunny just keeps running it's arrow up and down ad infinitum
[19:15] <toad_> you did switch the node off first?
[19:15] <iip_i2p> <Newsbyte> yes, I did
[19:15] <toad_> is there anything running according to task manager?
[19:16] <toad_> Couldn't retrieve key: SSK@rEWkqAYFXJrBs8DEAdI5RHLi0XQPAgM/FBNI//
[19:16] <toad_> Hops To Live: 15
[19:16] <toad_> Error: Route Not Found
[19:16] <toad_> Attempts were made to contact 3 nodes.
[19:16] <toad_> * 0 were totally unreachable.
[19:16] <toad_> * 0 restarted.
[19:16] <toad_> * 3 cleanly rejected.
[19:16] <toad_> * 23 backed off.
[19:16] <toad_> that's suspicious
[19:16] <iip_i2p> <Newsbyte> yes, there are 39 processes running
[19:16] <KenMan> 30 inbound connections in only 4 minutes uptime. Hmmm...
[19:16] * TheSeeker (~Fridlekh@cpe-65-28-158-211.socal.rr.com) has joined #Freenet
[19:16] <KenMan> they must like *me* :p
[19:17] <toad_> Newsbyte: is java running?
[19:17] <toad_> shit
[19:17] <iip_i2p> <Newsbyte> yes, and there lays the prob, I see
[19:17] <toad_> now that i've turned off all the logging I can't debug the proble,s
[19:17] <toad_> problems
[19:17] <iip_i2p> <Newsbyte> it's at 99 CPU
[19:18] <toad_> but if i turn the logging on, the node hangs
[19:18] <toad_> or drastically slows down
[19:18] <toad_> and i can only blame it on the logging because i can't see any other reason for it
[19:18] <toad_> grrrrrrrrrrrrrrrrrrrrrrrr
[19:18] <toad_> GRRRRRRRRRRRRRRRRRRRRR
[19:19] <iip_i2p> <Newsbyte> ok, killed and restarted, icon turned blue this time
[19:22] <iip_i2p> <Newsbyte> ok, killed and restarted, icon turned blue this time
[19:22] <jay> change the name of this channel to the Newsbyte support room
[19:23] <iip_i2p> <Newsbyte> my! What an improvement with the reseeding! Instead of 7 connections, I have two, now!
[19:23] * jay notices Fred generated a Sig 11
[19:23] <toad_> Newsbyte: instantly?
[19:23] <toad_> one minute after reseeding?
[19:23] <toad_> that's probably a good sign
[19:23] <toad_> I have 17, just after restarting
[19:24] <verl> oom at startup after reseeding - known problem?
[19:24] <toad_> verl: no
[19:24] <iip_i2p> <Newsbyte> it's a good sign not to have connections? Well, that promises, then.
[19:24] <KenMan> although I am not fully supportive of using a fixed keysize, as the vast majority are of the largest size, the wastage of using a big key (128 or 256) for a small file wouldn't be very much (percentage wise)...
[19:24] <toad_> unless you have set it to very low memory
[19:24] <verl> what logmessage?
[19:24] <verl> *want
[19:24] <toad_> Newsbyte: no, it's a good sign to have some connections
[19:24] <toad_> KenMan: huh?
[19:25] <toad_> the wastage of using a 128kB key for a 5kB file would be huuuge
[19:25] <iip_i2p> <Newsbyte> toad, I had *seven* a moment earlier, before I reseeded...
[19:25] <KenMan> most items transferred are 1M keys. So if a few 4K's get rounded up to 128, no big deal.
[19:25] <toad_> Newsbyte: yes, and it takes time to connect
[19:25] <toad_> KenMan: it'd be a big deal for the same reason that StoreData's are a big deal
[19:25] <toad_> as thelema said earlir
[19:25] <KenMan> by volume, very few items would be wasting a significant percentage of that 128 size...
[19:25] <toad_> earlier
[19:25] <toad_> if you're browsing fproxy, you don't want each small image to produce a massive transfer
[19:26] <toad_> woah
[19:26] <toad_> i have 49 now
[19:26] <iip_i2p> <Newsbyte> ok, in a nutshell; if I had 7, and after reseeding I have two, what particular benefit did reseeding give me, then?
[19:26] <KenMan> well, going by the numbers, the impact won't be that great.
[19:26] <toad_> Newsbyte: how many do you have now?
[19:26] <toad_> and why do you always expect instant results?
[19:26] <toad_> 51...
[19:26] <wind789> what about people on dial up?
[19:26] <KenMan> because the blue bunny looks so authoritative ?
[19:26] <toad_> I have goodish connections, do you want me to send you a locally generated seednodes?
[19:26] <iip_i2p> <Newsbyte> 3
[19:27] <toad_> 53...
[19:27] <toad_> w/ 6 minutes uptime...
[19:28] <verl> node won't start, oom every time, must be seednodes since it worked before i reseeded?
[19:28] <salahx> mayeb a corrupted seednodes file ?
[19:28] <toad_> Newsbyte: you want instant results, use Kazaa. I'm pretty sure you'll get 7 conns soon though.
[19:28] <toad_> verl: strange
[19:28] <toad_> at what point does it oom?
[19:28] <iip_i2p> <Newsbyte> 'instant'...geez, I've been running freenodes since more then 4 years, and I still didn't see one that really worked like it should, so I wouldn't call that 'instantly'. the 'your node just isn't connected well yet, give it time' is something I heard for the 5085 build too. But giving it time there didn't help.
[19:29] <verl> toad: i can send you my logfile?
[19:29] <toad_> verl: sure
[19:29] <wind789> newsbyte: you on stable?
[19:29] <iip_i2p> <Newsbyte> now, yes
[19:30] <wind789> I'm on unstable at the moment, have you tried the new unstable?
[19:30] <wind789> unstable is working fine for me personally
[19:31] <wind789> I guess I should switch back to stable, my node will get slammed either way
[19:31] <toad_> verl: it just says "got offer"
[19:31] <toad_> normally it would just accept it?
[19:31] <iip_i2p> <Newsbyte> I know, unstable usually works better then stable
[19:31] <toad_> it may be a problem on your end
[19:31] <iip_i2p> <Newsbyte> but at the end, it's supposed to be stable that works
[19:31] <wind789> I'm switching to stable...
[19:31] <verl> toad: http://w1.192.telia.com/~u19214808/freenet.log
[19:32] <jay> Newsbyte: you running Windows ? which version ?
[19:32] <toad_> verl: doesn't tell me very much :|
[19:32] <toad_> try logLevel=minor and do it again
[19:32] * KenMan is on stable for the long haul, at least beyond 'everything here is broken right now' ...
[19:32] <verl> toad: will do
[19:33] <iip_i2p> <Newsbyte> win3.11
[19:33] <iip_i2p> <Newsbyte> j/k ;-) win2000
[19:33] <KenMan> oh nice
[19:33] <jay> Newbyte: wow i was about to blow my stack
[19:33] <jay> eheh
[19:33] <jay> Win2K is good
[19:33] <mikeDOTd> i just upgraded to 5087 with the new freenet-ext.jar, and received an OOM on start
[19:33] <jay> Newsbyte: latest service pack?
[19:33] <iip_i2p> <Newsbyte> lol
[19:34] <verl> toad: um, now it started up
[19:34] <KenMan> i laugh at service packs too
[19:34] <toad_> you too?
[19:34] <toad_> strange
[19:34] <salahx> iuts it the native binary that's doing it ?
[19:34] <toad_> i'm sorry i have no idea why you get the oom
[19:34] <salahx> klaus@linux_duron:~/freenet > INFO: Native CPUID library jcpuid loaded from
[19:34] <salahx> resource
[19:34] <salahx> INFO: Optimized native BigInteger library 'libjbigi-linux-athlon.so' loaded
[19:34] <salahx> from resource
[19:34] <salahx> Caught java.lang.OutOfMemoryError running or seeding node
[19:35] <jay> KenMan: heh
[19:35] <verl> mikeDOTd: switch to loglevel minor and you can start :)
[19:35] <toad_> argh
[19:35] <jay> Newsbyte: do you know which service pack your 2K install is up to?
[19:35] <wind789> is the stable seednodes.ref file big enough?
[19:35] <toad_> lots of people getting the OOMs...
[19:35] <iip_i2p> <Newsbyte> hmm...going to 6...maybe it'll work after all...
[19:35] <mikeDOTd> verl: i'm trying with -Xmx256m now, i'll try that next thanks
[19:35] <iip_i2p> <Newsbyte> it's all the latest, rest assured
[19:35] * toad_ wonders why his computer freezes every so often... when it's frozen, it doesn't read the keyboard. it doesn't queue the keys, often it just loses them.
[19:36] <toad_> and then it comes back after a second or so
[19:36] * KenMan has the same malady !! :o
[19:36] <jay> toad_: that happens to me occasionally
[19:36] <lolo-laptop> toad_: mine does that too
[19:36] * toad_ suspects something dodgy with USB...
[19:36] <lolo-laptop> I think it has to do with dodgy PS2 touch pad on this cpu
[19:36] <toad_> jay, KenMan, lolo-laptop - you're all on linux 2.6, right?
[19:36] <lolo-laptop> yah
[19:36] <KenMan> mine does it for up to 200 seconds. But a ping on the internal interface always brings it right back. Check on the 2.6
[19:36] <jay> toad_: 2.4
[19:37] <lolo-laptop> hmm... mine is never more than 1 second
[19:37] <toad_> jay: hrrm, so it's a hardware problem... hrrm
[19:37] <iip_i2p> <Newsbyte> ok, 7, we're back at before we reseeded :-)
[19:37] <lolo-laptop> I sometimes get repeated keys to match keystrokes during the pause, any similar?
[19:37] <lolo-laptop> also, mine loses time, any similar?
[19:37] <KenMan> not here
[19:37] <KenMan> YES
[19:37] <KenMan> mine gains time
[19:37] <jay> toad_: it happened to me under 2.6
[19:38] <KenMan> for fun, my host likes to jump an hour into the future, and then come right back. Fred don't like that.
[19:38] <mikeDOTd> increasing the memory usage appears to have fixed the OOM on startup
[19:38] <toad_> so it might be a 2.6 problem...
[19:38] <toad_> mikeDOTd: what was the memory setting?
[19:38] <KenMan> These two issues are motivating me to move to defora...
[19:38] <mikeDOTd> it was 128m, now it's 256m
[19:39] <toad_> I don't see why the 5087 changes would increase the memory usage significantly...
[19:39] <toad_> i suppose it's the cached hexstrings on all the bigints?
[19:39] <KenMan> lolo-laptop - if you grep syslog for 'time reset' from ntpd, do you get any hits ?
[19:40] * toad_ recommends that everyone post (individually) their OOMs on support
[19:40] <lolo-laptop> my ntpd isn't working right...
[19:40] <toad_> I will see what I can do about them
[19:40] <toad_> tomorrow
[19:40] <toad_> bbl
[19:40] <KenMan> night toad, rest well
[19:40] <toad_> Newsbyte: what's it up to now?
[19:40] <toad_> I will talk to you tomorrow, buut tell me now
[19:41] * KenMan bets on 5
[19:41] <iip_i2p> <Newsbyte> still 7
[19:41] <iip_i2p> <Newsbyte> what, you are bbl'ing already? ;-)
[19:44] <KenMan> he says it, but he never really means it. Wait for the zzz part.
[19:44] <KenMan> :)
[19:44] <KenMan> goodness, he shouldn't go to sleep so soon after dinner ! Or maybe he is going off to have dinner now...
[19:45] <salahx> I can't replicate it on 60172
[19:46] <KenMan> lolo-laptop tell me 'bout your woes wid NTP ... please.
[19:46] <KenMan> I set up another host just so that I could ping this one on the internal NIC. That keeps the lockups down to a second or so :(
[19:46] <KenMan> And the soundcard loops on the last 400ms of played music, over and over and over and ... it bothers me.
[19:46] <KenMan> These incidents have increased in frequency over the past 2 months or more. Letzsee, I run 2.6.2 . And gave up running NTP when fred runs...
[19:46] <salahx> i ahve never had this problem
[19:46] * piranha (piranha@3ffe:b80:1ca1:0:0:0:deca:fbad) Quit (zelazny.freenode.net irc.freenode.net)
[19:46] * dystopia (admin@gaia-wan0.ipv6.freeshell.bofx.net) Quit (zelazny.freenode.net irc.freenode.net)
[19:46] * bluephile (bluephile@69-160-205-193.clvdoh.adelphia.net) Quit (zelazny.freenode.net irc.freenode.net)
[19:46] * thelema|aikido (~thelema@adsl-65-66-61-223.dsl.stlsmo.swbell.net) Quit (zelazny.freenode.net irc.freenode.net)
[19:46] <salahx> and I'm runing with NPTL !
[19:46] <lolo-laptop> KenMan: I just don't have it configured right and havne't had the time to get it right
[19:46] <KenMan> ah , good enough.
[19:47] <KenMan> I enabled stats logging for NTP. What i found was that every so often, the offsets would measure as 100's of seconds.
[19:48] * lostlogic_ (~lostlogic@node-402418b2.mdw.onnet.us.uu.net) Quit (zelazny.freenode.net irc.freenode.net)
[19:48] <KenMan> I finally decided it was my host to blame. It can't seem to keep the clock straight for about 1 in a million (or a thousand) gettimeofday()'s ...
[19:49] <lolo-laptop> KenMan: heh, ouch... mine just slowly loses time ... in accordance to how long thoses freezes last...
[19:51] <KenMan> I lose half a minute per day due to poor quality clock circuit, or something. If I don't run NTP, that is. I am long overdue for a system upgrade. I guess.
[19:51] <toad_> Newsbyte: how many now?
[19:51] <KenMan> I blame it all on software, so I am due for a software upgrade.
[19:51] <KenMan> other than the slow clock, that is. If I can run NTP, i don't really care how poorly uncalibrated my hard-clock is...
[19:53] <KenMan> shhh, you'll wake the toad. Oops, I mean...
[19:53] <KenMan> if he says >=10 will you type zzz ?
[19:53] * thelema|aikido (~thelema@adsl-65-66-61-223.dsl.stlsmo.swbell.net) has joined #freenet
[19:53] * bluephile (bluephile@69-160-205-193.clvdoh.adelphia.net) has joined #freenet
[19:54] <KenMan> these are weird net-splits...
[19:54] <salahx> Well freenet is over capacitiy it loooks like
[19:55] <salahx> err freenode rather
[19:55] <salahx> Freenet is ALWAYS over capacity :)
[19:55] <jay> kornbluth went down
[19:55] <KenMan> must just be the governmental uber-hackers out to get us 8^) i saw trouble with orwell just now...
[19:55] * piranha (piranha@3ffe:b80:1ca1:0:0:0:deca:fbad) has joined #freenet
[19:55] * dystopia (admin@gaia-wan0.ipv6.freeshell.bofx.net) has joined #freenet
[19:56] * KenMan confesses he knows zilch about managing an irc network such as freenode...
[19:56] <toad_> Newsbyte is hiding because he lost the argument :)
[19:56] <jay> hehe
[19:56] <KenMan> that's what I think...
[19:57] <KenMan> or maybe he went to the #newsbyte-support channel ! :o
[19:57] <iip_i2p> <FillaMent> I'd just like to note that I haven't seen persistant RNFs in over a month, since updating to 5087 I feel lucky to get a DNF
[19:58] <toad_> FillaMent: IMHO the network has been through a bad patch, but it will heal
[19:58] <toad_> also it needs to adjust to the new routing
[19:58] <toad_> bbl again :)
[19:58] <salahx> ok
[19:58] * KenMan notes the lack of a zzz :)
[19:58] <iip_i2p> <FillaMent> I'll take your word for it. Hitherto, I've just been enjoying the fact that other people can't get crap and I'm doing fine
[19:59] <iip_i2p> <Newsbyte> luckily, toad isn't bbl'ed ;-)
[20:00] * KenMan likes to fill in the dead time with some mindless one-way chatter ...
[20:00] <KenMan> So I took one of my family's dogs for a swim the other week, and it made her fur really fluffy. She is still shedding the undercoat as a result...
[20:01] <KenMan> We went swimming in a river when the water was flowing strong, and the bottom was all big rocks :(
[20:01] <KenMan> she kept getting washed away (70 lbs) so I had to drag her back everytime she swam through the stongest channel...
[20:02] <iip_i2p> <Newsbyte> make a flog about it, kenman ;-)
[20:02] <KenMan> my sister bathes her dogs in a full tub, and does a decent job. My mother washes her pooch in the shower, and can't do a very good job.
[20:02] * KenMan is just filling in the "dead-air"
[20:03] * TheSeeker (~Fridlekh@cpe-65-28-158-211.socal.rr.com) Quit (Read error: 110 (Connection timed out))
[20:03] <KenMan> Sunday I filled a baby pool for my mother's dog, and she wouldn't get into it. It was a hot day and I knew it would cool her off. But she is older now, and didn't want to be bothered...
[20:03] <iip_i2p> <Newsbyte> on a positive sidenote, I managed to get FIND
[20:03] <KenMan> Eventually we forced her in, to give her a bath...
[20:04] <iip_i2p> <Newsbyte> didn't happen with the 5085
[20:04] <KenMan> but "my" doggies love me, because I walk them and give them all sorts of good food, and attention.
[20:05] <KenMan> The best part used to be that I could go home, and didn't have to do the maintenance. But lately I'm picking up more than my fair share of poops.
[20:05] <jay> 20 RNF's in a row
[20:06] * lostlogic_ (~lostlogic@node-402418b2.mdw.onnet.us.uu.net) has joined #freenet
[20:06] <iip_i2p> <Newsbyte> heh...well, I have a strong opinion on doggie-issues
[20:07] <KenMan> yeah, cats just piss on things so badly that you can't wash out the smell...
[20:07] <salahx> Still be nice to see a libjbigi-linux-dynamic.so in the jar though
[20:07] <KenMan> wait, jay was the cat lover , dog hater...
[20:07] <jay> yeah
[20:07] <KenMan> :)
[20:08] <jay> i refuse to date a chick with a dog
[20:08] <jay> now i will get a dog if i have a house and a yard for em
[20:08] <KenMan> i refuse to tolerate a chick's cats.
[20:08] <jay> but female dog owners can't train dogs
[20:08] <jay> not the ones ive dated anyhow
[20:08] <KenMan> yes.
[20:09] <iip_i2p> <Newsbyte> I've seen a documentary about USA women in Florida (and my mom told me recently a similar story about japan) where old, aparently degenerated 'ladies' filled there days with powdering and parfuming their little fluffy rats..ermm...dogs...going to the 'massage' with them, spending mink-coats of 8000 bucks to clothed their little darling, etc...
[20:09] <KenMan> some just don't know you can do that. "Your dog is humping my leg" ... "oh, he only does that to people he likes, <giggle>"
[20:09] <iip_i2p> <Newsbyte> in short, it made me puke
[20:09] <jay> every five minutes it's "No, don't do that", "<name-of-dog> stop that!", etc...
[20:09] <KenMan> heh, yeah. that's the other annoyance.
[20:09] <jay> KenMan: heh
[20:09] <jay> oh it's torture
[20:09] <jay> massive turn-off
[20:10] <jay> watching a woman who can't manage the dog while it's licking my arms
[20:10] <KenMan> my sister has whipped all her dogs into very obedient creatures. One even learned the command "wag your tail" extremely well.
[20:10] <jay> girls can do it.. most just choose to "let the dog be free" to piss everywhere
[20:10] <iip_i2p> <Newsbyte> sigh
[20:11] <KenMan> Newsbyte knows that if he shares his number of connections, the toad will leap to the zzz lilypad...
[20:11] <toad_> KenMan: motherboard?
[20:11] <toad_> jay: motherboard?
[20:12] <jay> toad_: what did u call my mother>?
[20:12] <iip_i2p> <Newsbyte> well, tastes differ, I guess. to begin with, for me, a dog should be a real dog, like a german sheppard, not those special breeded rats that are no biger then a well-fed spider. I mean, step on such a dog, and it's dead. gezz.
[20:12] <jay> i have a Dragon KT600 Plus!
[20:12] <jay> soyo
[20:12] <jay> Newsbyte: i second that
[20:12] <KenMan> toad is asking if we engage in unnatural practices. I don't do that one.
[20:12] <jay> a dog shouldn't look like a rat on steroids
[20:12] <KenMan> my troublesome computer has an EPoX motherboard.
[20:13] <jay> KenMan: did you build the cpu yourself using scripts?
[20:13] <KenMan> it was never troublesome until after 2.6 though. 3 years of solid service.
[20:13] <KenMan> Yes, I did.
[20:13] <jay> KenMan 2.5G
[20:13] <toad_> KenMan: chipset?
[20:13] <jay> make sure there's a cpuid for it
[20:13] * lolo-laptop (~lostlogic@68.251.84.226) Quit ("Client exiting")
[20:13] <KenMan> only I know how translate scripts into silicon. And it will stay that way - hahaha !!
[20:13] <toad_> yeah, it's prolly 2.6...
[20:14] <iip_i2p> <Newsbyte> I'm engaed in unnatural practices..
[20:14] <jay> 2.6 is real buggy
[20:14] <jay> imo
[20:14] <jay> i went back to 2.4 because of the strangeness in 2.6
[20:14] <toad_> it's good for freenet...
[20:14] <KenMan> chipset is ... um ... VT8366/A/7 [Apollo KT266/A/333] - VIA
[20:14] <iip_i2p> <Newsbyte> I'm trying to run a freenode and hope it will do well..what's more unnatural then that? ;-)
[20:14] <toad_> hmmm
[20:14] <toad_> okay so we're all on VIA chipsets
[20:14] <toad_> but some variety
[20:15] <toad_> hrrm
[20:15] <jay> toad_: yeah i got the 8237
[20:15] <toad_> VIA for socket A, that's about all that's in common...
[20:15] <toad_> jay: aka kt600?
[20:15] <toad_> yeah
[20:15] * toad_ is kt400
[20:15] <jay> toad_: yup
[20:15] <toad_> oh well
[20:15] <iip_i2p> <Newsbyte> ermm...Aopen?
[20:15] <toad_> probably buggy driver or something
[20:15] <toad_> bbl
[20:15] <KenMan> I've had 2+ years of solid service from this box. less than 6 months of crap, that keeps becoming more frequent...
[20:16] <jay> i think some of VIA's chipsets are shitty
[20:16] <KenMan> i am kt266A i think . I think some of VIA's chipsets have some shitty problems that will never be resolved.
[20:17] <jay> mine generates 'spurious interrupts on IRQ 5 or 7' randomly
[20:17] <KenMan> I will jump to Fedora, and then everything will be stable again. Or else I will move to a newer box.
[20:17] <jay> nobody on the kernel list really knows why, other than via is buggy
[20:17] <iip_i2p> <Newsbyte> well, it's late, I'm off
[20:17] <jay> oof fedora
[20:17] <jay> run debian
[20:17] <KenMan> those spurious interrupts are supposedly 'safely ignored'
[20:17] <jay> KenMan: yeah i removed the printk from the kernel source
[20:18] <KenMan> hee hee, I did that for someone else ...
[20:18] <jay> it was going right to the console screwing up whatever i was working on
[20:18] <KenMan> just recently too
[20:18] <KenMan> yup. Mine was for a non-computer-savvy user, and something like that might scare them.
[20:18] <KenMan> Later, her mother told me, "no, she would just ignore something like that" ...
[20:19] <jay> good ol end user
[20:19] <KenMan> so i wasted 15 minutes finding out that it was already going to be 'safely ignored' ;)
[20:19] <jay> you don't see it in GUI mode anyway.. just console if i recall correctly
[20:20] <KenMan> come to think of it, my "kernel patch" didn't make it onto the final CD i burned for her. Which was just console only, for laptop backup and restores...
[20:20] <KenMan> so i made the fix, and felt smug in the knowledge that I could do it. Oh well...
[20:20] <jay> my JVM is now consistantly crashing on Sig 11's for either inserts or just normal operation
[20:21] <KenMan> at first I thought "bad hardware" associated with 11. Then I got it one time, and said "no it couldn't be that!" 8)
[20:21] <KenMan> 1.4.2_03 with 5085 or 5086...
[20:21] <jay> the same PC address the last 3 times
[20:22] <KenMan> odd !!
[20:22] <jay> Function=(null)+0x4031F48A
[20:22] <jay> the last 3 times
[20:22] <jay> wherever the f*ck that is
[20:22] <KenMan> perhaps it was a stack overwrite ?
[20:23] <KenMan> naw, java could NEVER do something like that... heh
[20:23] <jay> yeah it's a safe language.. it never crashes
[20:23] <jay> garbage collection fixes all that right?
[20:23] <KenMan> we've seen it with several JVMs... of course a "sig 11' is only generated on a linux box...
[20:23] <jay> oh and there's no pointers too.. whew
[20:23] <jay> happened on 2.6 and 2.4
[20:23] <KenMan> :(
[20:24] <KenMan> java is trying to get us back for switching over to non-java big integers...
[20:25] <jay> i think that's what's causing it.. not that it's incorrect, but that it's going down a codepath generating Sig 11's on linux
[20:25] <KenMan> actually it's an early test of some microsoft DRM initiative, secretly inserted into the JVM...
[20:25] <KenMan> after that MS-Sun deal...
[20:25] <jay> i would have loved to been in that meeting
[20:25] <KenMan> "use of unauthorized big integers" ...
[20:25] <jay> heh
[20:26] <jay> maybe we need to sign an NDA with ms to use big ints
[20:26] <KenMan> actually, it could be related to our use of JNI, perhaps...
[20:26] <jay> that too
[20:26] <jay> it's certainly something very recent
[20:26] <KenMan> we hit the JNI path pretty hard.
[20:26] <KenMan> with short execution times...
[20:26] <jay> when we were testing unstable and doing the network reset before distributing the new stable
[20:27] <KenMan> yup
[20:27] * Ash-Fox (Hal-9000@aae32.neoplus.adsl.tpnet.pl) Quit (Read error: 104 (Connection reset by peer))
[20:27] <jay> im going to violate a certain software license.. brb
[20:27] <KenMan> this is getting ridiculous - lockups every 5 minutes or so... now where did I put that "desktop tazer" ??? time to teach this box a lesson.
[20:28] <salahx> Here's a good expample and auothrized BigInt: http://primes.utm.edu/curios/page.php?number_id=953 :)
[20:29] <KenMan> funny
[20:29] * Ash-Fox (Hal-9000@aae32.neoplus.adsl.tpnet.pl) has joined #FreeNET
[20:29] <jay> im going to patent the prime number 13
[20:30] <jay> i really invented it
[20:30] <KenMan> you better hurry, or else you will have bad luck
[20:30] <jay> that patent expires
[20:30] <jay> expired even
[20:30] <jay> now it's a good number
[20:31] <jay> i upgraded debian over the weekend and now the output for 'ls -l' is much narrower than before...
[20:31] <jay> strange
[20:31] <salahx> maybe an alais is gettign in teh way ?
[20:32] <jay> nah i checked that
[20:32] <jay> ll is aliased to ls -l anyway
[20:32] <KenMan> so did /bin/ls get modified ?
[20:33] <KenMan> can aliases be recursive ??
[20:33] <jay> -rwxr-xr-x 1 root root 75948 Jul 16 07:37 /bin/ls
[20:33] <jay> i guess so
[20:34] <KenMan> does anybody have the real time ? my box says 8:45...
[20:34] <jay> Wed Jul 28 00:52:26 UTC 2004
[20:34] <KenMan> tank you, apparently i've fallen to sanity's level of laziness !! :o
[20:35] <jay> you could CTCP me
[20:35] <KenMan> or i could just teach this machine a lesson it will soon forget... :)
[20:35] <jay> with the tazer
[20:35] * KenMan fumbles around for the tazer
[20:36] * KenMan (~chaziller@pcp403292pcs.mntcrm01.md.comcast.net) Quit ("ChatZilla 0.9.61 [Mozilla rv:1.7b/20040324]")
[20:36] <jay> haha
[20:36] <jay> chatzilla.. the all time worst name for a chat client
[20:37] * jay goes to code C
[20:41] <wind789> my upload limit of 30 kb/sec is doing a pretty good job of holding it right at 37 kb/sec
[20:41] <wind789> I tried 40 and it was fluctualting all over the place
[20:59] * salahx (salahx@sc1-24.217.174.147.charter-stl.com) Quit ()
[21:02] * Ash-Fox (Hal-9000@aae32.neoplus.adsl.tpnet.pl) Quit (Client Quit)
[21:05] * Ash-Fox (Hal-9000@aae32.neoplus.adsl.tpnet.pl) has joined #FreeNET
[21:30] * jay (jcl@ool-18bf6dac.dyn.optonline.net) Quit ("Client exiting")
[21:31] <toad_> lol, that's funny... http://decss.zoy.org/ explains how he doesn't care about getting decss over freenet because freenet is only for piracy whereas decss has legitimate uses. the punchline: the FTP servers for downloading the decss movie are not available, probably due to bandwidth issues
[21:37] * salahx (salahx@sc1-24.217.174.147.charter-stl.com) has joined #freenet
[21:42] * TheSeeker (~Fridlekh@cpe-65-28-158-211.socal.rr.com) has joined #Freenet
[22:03] <wind789> hehe
[22:20] <wind789> I wonder if freenet nodes will ever earn priority credits from each other ?
[22:21] * jmercier (~jmercier@4.10.164.111) has joined #freenet
[22:22] * TheSeeker (~Fridlekh@cpe-65-28-158-211.socal.rr.com) Quit (Read error: 110 (Connection timed out))
[22:24] <salahx> priority credits ?
[22:25] <wind789> some p2p systems do it
[22:25] <wind789> reward good nodes
[22:25] <wind789> I'm not sure what you'd reward though
[22:25] <salahx> FReenet works the other way around
[22:25] <wind789> punish bad nodes?
[22:25] <salahx> FReent uses "negative trust" i.e. in punshes bad nodes
[22:26] <salahx> other system use "positive trust" i.e. reawrd good nodes
[22:26] <wind789> actuly since you mention it, other systems use both
[22:26] <wind789> ban people that don't share ect
[22:28] <wind789> there needs to be a way to deal with people that set the upload limit at vitually nothing and then download movies with 30 threads
[22:28] <wind789> I know, easier said than done....
[22:29] <salahx> its already dealth with
[22:29] <wind789> ?
[22:29] <salahx> those nodes get "punished" by the estimaros
[22:29] <salahx> errr by the estimators
[22:30] <wind789> what if they open 30 connection to 30 different nodes and download 30 chuncks at once? would they still be punished?
[22:30] <salahx> Besides in FReent you woudlnt; want to disable uplaoding - if you did, you'd lose all your plausible deniablilty
[22:30] <salahx> Requests are handle by the mRI
[22:31] <wind789> I wish the default upload limit didn't have to be so slow....
[22:31] <salahx> well msot people haver 128k up so that's what it was sent for
[22:31] <wind789> you'll have to forgive me, I'm not entirly informed about the internals of freenet, what is MRI?
[22:32] <salahx> mRI = Minimum REquest Interval
[22:32] <wind789> ah
[22:33] <wind789> I'm on a regular cable connection, I can upload up to about 60 kb/sec
[22:33] <salahx> typivcally speaking, you dowenlaod speed iwll be the same as your upload with FReenet
[22:33] <wind789> DSL is my area id much slow I think
[22:34] <wind789> hhmm
[22:35] <wind789> yeah but isn't the download speed reduced since it's using the upload resources of many people?
[22:36] <wind789> for instance if I download 1 meg at htl 5, it's thay the same as 5 people sending 1 meg?
[22:36] <wind789> isn't that the same*
[22:36] <salahx> assuming the lastn ode has it, then yes
[22:36] <wind789> hehe
[22:37] <salahx> IF that souinds inefficent; it is. Such measures are needed for annonymiity though
[22:38] <wind789> I try to upload many many times more than I download to try to keep from leeching
[22:38] <salahx> Leeching isn;t a problem on Freenet
[22:38] <wind789> but it doens't hurt to contribute...
[22:38] <salahx> wel of course
[22:38] <salahx> By all means yif you have good content insert it
[22:39] <wind789> not like that, I mean I keep my node running
[22:39] <salahx> but even peopel who insert nothign stil help becsaue any (well ,most due otr pcache)day that passed though there node get stored
[22:39] <wind789> it has a ton of files in the DS so it acts much more like a server than a proxy
[22:39] <salahx> yeah Freenet oen of thseo tghing you want to leaving running all tge time
[22:39] <salahx> it acts as both
[22:40] <salahx> and no one knows which one :)
[22:40] <wind789> my upload bandwidth is much higher than my download bandwdith
[22:40] <wind789> so I've got a lot more stuff leaving my node than entering it
[22:41] <wind789> if you request a key from sombody at htl 1 and they instantly send it to you, can you assume they had it in their DS?
[22:42] <salahx> no
[22:42] <salahx> becase the HTL is fudged
[22:42] <salahx> when your quests passes though a node, the HTL isn;t always decremed, to add uncertaity
[22:43] <salahx> the lower the HTL the high the probbolit its not decremeted
[22:43] <wind789> I was not aware of that
[22:43] <salahx> The forumla used by Freenet, as a result, add abotu 2 to 3 hops with HTL=20
[22:45] <wind789> if their DS is not full, can you assume that they now have the file in their DS weather they original had it or not?
[22:47] <salahx> if you insert it, yeah
[22:47] <salahx> But of couse, teh other person coudl lefitaitmy say "I was framed!"
[22:47] <wind789> lol yep
[22:47] <salahx> futhermore all the data in FReenet is encryped, unless you know the routing key, you have no idea what in your datastore
[22:49] <salahx> Teh actual details of how that works thoiguh are UGLY
[22:49] <salahx> ESsentiall is a hash of a hash of a hash.
[22:50] <wind789> assuming that you collected all the routing keys you could find, could keep tabs on what files are beeing most requested?
[22:50] <salahx> Sure, adn FReenet does exactluy that
[22:50] <salahx> the the datasotre is full adn it needs to make room, it throws out the least used keys
[22:51] <wind789> yeah but it doesn't actauly know what's in the file without the routing key correct?
[22:51] <salahx> Excatly
[22:51] <salahx> until it gets the request (which includes the exnrpytion key the node does not know), it has no idea
[22:52] <wind789> it would be neat if sombody wrote an add-on that would generate stats showing which rouing keys were most requested bases on their assocated DS keys
[22:52] <wind789> you'd have to know both keys of course
[22:52] <salahx> but you wouslbn;t WANT to
[22:52] <wind789> hehe I know, but it would be neat
[22:52] <salahx> that woud lbe hard becasue you c'anrt guratee a request wil lbe rotuer though you
[22:52] <wind789> you could have a free site showin the most popular knows files on freenet
[22:53] <wind789> true, it might give some rough ideas
[22:53] <salahx> The most popualr stiff is the stuff you cna actuaklly retrieve :()
[22:53] <wind789> lol
[22:53] <wind789> and vice versa
[22:54] <wind789> it seems unpoplar stuff(retrivable or not) dies off rather quickly
[22:54] <salahx> well nto really
[22:54] <salahx> its not that it vanisheds frmo ther datastore, but become unreachable due to (bad) routing
[22:55] <wind789> lol
[22:55] <salahx> When the devs "Reset" the netowrk, often stuff over 2 yers old is suddenly reachable again
[22:55] <wind789> wow, that's news to me
[22:55] <salahx> yep
[22:55] <salahx> the main proble is the data out there - it how to ind it (and keep it anyonomus) is the tough part
[22:56] <wind789> I suspected, and still do that many people have small data stoes so I bought a 120 gig hard drive for freenet
[22:56] <salahx> cool
[22:57] <salahx> actually most of the datasotre are probalby quie big - disk space is cheap and Frost needs it :)
[22:57] <wind789> when my store is emty it fills at a least a gig every 2 days
[22:57] <salahx> good
[22:57] <wind789> yeah but do people allocate much sapce?
[22:58] <wind789> how is the default DS size determined?
[22:58] <salahx> for the inwinstaller, 10% of the hard disk
[22:59] <salahx> For Linux users, 256M but they are rpompted to change it
[22:59] <wind789> lol cool
[23:00] <wind789> does probabalistac caching still exist?
[23:01] <salahx> yes
[23:01] <salahx> it onyl takes effect if your datastore is full though
[23:01] <wind789> I see
[23:04] <wind789> well I'm going afk, good talking to you, thanks for the info cya
[23:04] * Tailchaser (~tailchase@pool-68-162-16-41.nwrk.east.verizon.net) has joined #freenet
[23:07] <salahx> ok!
[23:33] * jmercier (~jmercier@4.10.164.111) Quit ("Leaving")
[23:42] * TheSeeker (~Fridlekh@cpe-65-28-158-211.socal.rr.com) has joined #Freenet
[23:57] * Tailchaser (~tailchase@pool-68-162-16-41.nwrk.east.verizon.net) Quit (Remote closed the connection)
These logs were automatically created by Jay Oliveri with his gimp hapi on irc.freenode.net.
