Timestamps are in GMT/BST.
[0:15] * Ash-Fox (Hal-9000@abh154.neoplus.adsl.tpnet.pl) Quit (Read error: 54 (Connection reset by peer))
[0:15] * Redb3ard (~oylerj@c-24-125-12-101.va.client2.attbi.com) Quit (Read error: 232 (Connection reset by peer))
[0:16] * Ash-Fox (Hal-9000@abh154.neoplus.adsl.tpnet.pl) has joined #FreeNET
[0:33] * Zorix (Brandon@fl-198-70-221-44.dyn.sprint-hsd.net) Quit (Read error: 110 (Connection timed out))
[0:40] * lolo-laptop (~lostlogic@node-4024215a.mdw.onnet.us.uu.net) Quit ("Client exiting")
[1:21] * Zorix (Brandon@fl-198-70-221-44.dyn.sprint-hsd.net) has joined #freenet
[1:35] * ShaunMacPherson (~ShaunMac@Sudbury-HSE-ppp3977356.sympatico.ca) has joined #Freenet
[3:10] * moskau23 (~Miranda@dsl-082-082-233-220.arcor-ip.net) has joined #freenet
[3:28] * Zorix (Brandon@fl-198-70-221-44.dyn.sprint-hsd.net) Quit (Read error: 110 (Connection timed out))
[3:38] * pittaman (~mistery_b@d51A410CA.kabel.telenet.be) Quit ("Fuck that shit yo")
[5:18] * sanity (~ian@81-178-94-146.dsl.pipex.com) has joined #freenet
[6:30] * dystopia (admin@gaia-wan0.ipv6.freeshell.bofx.net) Quit (Read error: 60 (Operation timed out))
[6:35] * kers (kers@m213-101-157-3.swipnet.se) has joined #freenet
[6:55] * hobx_ (~chatzilla@ankh.math.chalmers.se) Quit (Remote closed the connection)
[6:59] * kers (kers@m213-101-157-3.swipnet.se) Quit (Read error: 54 (Connection reset by peer))
[7:12] * dystopia (admin@gaia-wan0.ipv6.freeshell.bofx.net) has joined #freenet
[7:52] * Hory (~Miranda@81.196.25.110) has joined #FreeNet
[8:05] * guido^pe (~unknown@dsl-213-023-232-206.arcor-ip.net) has joined #freenet
[8:34] * cbreak (~cbreak@dhcp-32-094.via-eth.ch) has joined #freenet
[9:04] * Hory (~Miranda@81.196.25.110) Quit (Read error: 110 (Connection timed out))
[9:27] * Ash-Fox (Hal-9000@abh154.neoplus.adsl.tpnet.pl) Quit (Connection timed out)
[9:47] <toad_> whew
[9:48] * toad_ recommends somebody read his latest post about simulations...
[9:48] <toad_> might be an interesting read!
[9:48] <toad_> especially ian, if he's not asleep as usual
[9:57] * FallingBuzzard (~srademach@207.152.112.129) has joined #freenet
[9:57] * kers (~kers@h79n2fls33o847.telia.com) has joined #freenet
[10:02] * lolo-laptop (~lostlogic@68.251.84.226) has joined #freenet
[10:06] * jay (jcl@ool-18bf6dac.dyn.optonline.net) has joined #freenet
[10:09] <toad_> hi people
[10:09] <toad_> hi jay
[10:09] <toad_> are you here?
[10:15] <jay> hello
[10:16] <jay> sup frog?
[10:16] <toad_> read my recent post to tech yet?
[10:17] <toad_> Subject: [Tech] Latest simulation results - apparently polynomial scaling
[10:17] <toad_> i will probably need to talk to ian about it...
[10:18] <jay> i just read it
[10:19] <toad_> thoughts? comments? ideas? wild ravings?
[10:19] <toad_> prophecy? encouragement? despair? bullets?
[10:20] <jay> lol
[10:20] * sanity (~ian@81-178-94-146.dsl.pipex.com) Quit ()
[10:20] <jay> well NGR apparantly can work
[10:20] <jay> ie it isn't broken
[10:20] <toad_> but not scale
[10:21] <toad_> if it doesn't scale then for our purposes it IS broken
[10:21] * guido^pe (~unknown@dsl-213-023-232-206.arcor-ip.net) Quit (Remote closed the connection)
[10:21] <toad_> we need to be able to support networks of 10k nodes at least - firstly because that's what we've got, secondly because that's arguably the minimum for anonymity
[10:22] <jay> i mean that if it works it may still be fundamentally broken.. whatever that means
[10:22] <toad_> In any case judging by the above progression, any really large
[10:22] <toad_> network would have to have an impractically large HTL to achieve the
[10:22] <toad_> targetted result.
[10:23] <toad_> combined with the fact that I'm told that the original classical routing simulations aren't as persuasive as they are widely reputed to be...
[10:23] <cbreak> so what is a solution? A New Routing Algorithm? Fixed Size Keys make things faster, including learning, but don't improve routing.
[10:24] <toad_> well firstly more simulations to find out if this pattern replicates with a lower target psuccess
[10:24] <toad_> but if it does...
[10:24] <jay> the problem with NGR is that the state information it holds is always (in a relative way) out of date
[10:24] <toad_> jay: hmm?
[10:24] <toad_> it might also be an issue with announcements
[10:24] <jay> akin to split brain.. ie different nodes have different information about each other
[10:25] <toad_> it's just possible that the psuccess will recover and rise to a completely different level...
[10:25] <toad_> but judging by some of the graphs, that'll take an insane amount of time and doesn't seem likely
[10:25] <jay> i had hoped that even with this NGR could achieve some success
[10:25] <toad_> jay: it's the same with classical routing
[10:26] <toad_> jay: we always have limited info on nodes, and usually different nodes have different info on us
[10:26] <jay> classical routing made more or less a random choice, yes?
[10:26] <toad_> the only way we could have equal info is a classical DHT
[10:26] <toad_> jay: no
[10:26] <jay> random first choice?
[10:26] <toad_> classical routing routed to the node whose reference was closest to the key we are looking for
[10:26] <toad_> first hop was random of course
[10:27] <jay> the fixed key size idea is interesting
[10:27] <toad_> jay: and irrelevant
[10:27] <jay> i like it, but id like it to happen smoothly instead of in a rush
[10:27] <toad_> if routing doesn't scale, nothing else matters
[10:27] <jay> not according to some of your observations prior
[10:27] <jay> ah right
[10:28] <jay> well routing (i thought) was complicated by different sized keys over different bw's
[10:28] <toad_> yes, in reality it is
[10:28] <toad_> but in simulation we take out most of the complicating factors
[10:28] <toad_> we can put them back in if we find an algo that scales
[10:29] <jay> ok i see.. so those weren't taken into account these runs
[10:29] <toad_> right
[10:29] <toad_> it assumes fixed file size
[10:29] <jay> if it don't work in a perfect world.. then i see your point
[10:29] <toad_> and transfer time is simply the number of hops from the source
[10:29] <toad_> if it doesn't scale in a perfect world... we have major problems...
[10:29] <jay> i meant scale actually
[10:30] <jay> toad_: how much more accurate would you say these sims are compared to the serapis one i
[10:30] <jay> 'm aware of
[10:30] <toad_> jay: the serapis sims were of classical routing
[10:31] <toad_> jay: what i heard about them was that the graph can be read either way
[10:31] <toad_> even if the graph is clearly logarithmic, we still have the eternal headache of how to make them work with load...
[10:31] <jay> yes and that's a major headache
[10:31] <toad_> but even if that is true, abandoning NGR is a little concerning!
[10:32] <toad_> freenet is far less interesting with classical routing
[10:32] <toad_> in terms of deployment possibilities, attack resistance, etc
[10:32] <toad_> having said that, we've already debunked the idea of trusted links only steg in hostile environments - you need a significant connection flux to maintain logarithmic HTL, according to kleinberg...
[10:33] <jay> is ian against dht's?
[10:33] <toad_> there may be ways around that though..
[10:33] <toad_> they'd look like traditional DHTs though
[10:33] <jay> or is it just a lot of work that won't solve everything?
[10:33] <toad_> jay: ian believes both classical and ng routing are some form of heuristic DHT
[10:33] <toad_> he also points out that getting a content distribution DHT working is hard
[10:33] <jay> i like the dht idea in concept
[10:34] <toad_> provable properties are nice, admittedly
[10:34] <toad_> but not necessarily especially attack resistant
[10:35] <toad_> I2P is harvestable, for example... but it looks like freenet will always be harvestable too...
[10:36] <jay> i've been thinking that, in general, a freenet that doesn't "try too hard" to deliver messages would be more scalable (taking from hobx's ideas)
[10:37] <toad_> i'm not simulating retries
[10:37] <toad_> or load
[10:37] <jay> ok sticking with the sim\
[10:37] <jay> do u have the old png's online somewhere?
[10:38] <toad_> no
[10:38] <toad_> some of them are in the various papers though
[10:39] <jay> im actually concerned with the 1.4/1.5 debate and newsbyte's complete ignorance
[10:39] <jay> but glad with your comments on the matter
[10:40] <jay> ie recommend 1.5, use language features no later than 1.4 until further review
[10:40] <toad_> page 8 of the freenet-ieee paper
[10:40] <toad_> shows O(n^0.28)
[10:40] <toad_> looks like a pretty good match to me
[10:40] <toad_> ian says you can match a log to it
[10:40] <toad_> i'm not convinced
[10:42] <toad_> i'm very reluctant to significantly increase the max HTL
[10:43] <jay> i would be too
[10:43] <toad_> not much point trying to build an attack resistant network if you have a work factor of 100!!
[10:43] <jay> is it at 20 currently?
[10:43] <toad_> yes
[10:43] <cbreak> how does the max htl affect the average htl?
[10:43] <toad_> xn
[10:43] <toad_> cbreak: debatable
[10:43] <toad_> on 0.3, the average htl slowly rose to the max htl
[10:43] <toad_> but that was without pcaching
[10:44] <toad_> anyway i'm talking average here!
[10:44] <toad_> on the basis of these graphs, for a 1000 node network you need ~ 45
[10:44] <toad_> for a 10,000 node network you need rather more
[10:45] <toad_> average increase in log of 0.343 for each doubling
[10:46] <toad_> so if 800 is l(40)...
[10:46] <toad_> 10,000 is approximately 800 * 12 - lets call it 800 * 16, which would be 12,800 nodes
[10:47] <toad_> that makes an HTL of over 157...
[10:56] <toad_> bbiab
[11:02] <jay> toad_: why are some of the lines (ie nodes=200/htl=15) cut at some point (that one around 1M) ?
[11:03] <toad_> jay: because they run at different speeds
[11:03] <toad_> some of them i stopped to run others
[11:03] <jay> what kind of lines were you hoping to see?
[11:04] <jay> in general
[11:08] <jay> because i would expect the curve to start low and go high.. im missing something
[11:08] <jay> i just checked some of the older graphs from your emails
[11:08] <toad_> starts high, goes low
[11:08] <toad_> like the other announcement graphs
[11:08] <toad_> the reason is that we continually add more content
[11:09] <toad_> so there's a swamping effect
[11:09] <toad_> at the beginning there is little content in the network, we can centralize it a bit
[11:09] <toad_> later on that's not possible any more
[11:09] <toad_> due to small datastores
[11:09] <toad_> at least, that's my thoeyr
[11:09] <toad_> theory
[11:09] * i2p_iip (~changate@host.teitel.net) Quit (Remote closed the connection)
[11:09] <jay> ah ok.. makes sense
[11:09] <toad_> well i think so, but i'm probably completely wrong
[11:10] <jay> i wanted to know your thinking whether it's right or wrong :)
[11:11] <cbreak> how does ngr learn? From Inserts? Successfull requests? Or also from failures?
[11:11] <toad_> well i wrote it..
[11:11] <toad_> successful, and failed, requests
[11:11] <toad_> 3 estimators - pDNF, tDNF, tSuccess
[11:11] <toad_> combined: tSuccess*(1-pDNF) + tDNF/(1-pDNF)
[11:12] <jay> math confuses me
[11:13] <toad_> life confuses me
[11:13] <toad_> freenet too
[11:13] <jay> it's been 10 years since i worked with any math
[11:13] <jay> corporate programming doesn't usually require it
[11:13] <jay> but i understand your current graph
[11:14] <jay> and i have no real advice on the sim in particular
[11:14] <jay> the user perspective (from the frost boards) is currently focused on inserts.. it seems they can retrieve very well at the moment
[11:14] <toad_> well ian has passed judgement and apparently isn't interested in continuing the conversation...
[11:15] <toad_> well inserts at the moment are spammed a bit...
[11:15] <toad_> because of the timeout issues
[11:15] <jay> what is it that's timing out specifically?
[11:16] <jay> just got ian's response
[11:16] <toad_> inserts
[11:16] <jay> heh
[11:16] <toad_> too early
[11:16] <jay> no further info?
[11:16] <toad_> hmm?
[11:16] <toad_> the insert timeouts were all too short
[11:16] <cbreak> he mentioned OS X :) Cool :)
[11:16] <toad_> cbreak: he's an ex-linux mac user...
[11:17] <toad_> at least the kernel is open source :)
[11:17] <jay> brb..
[11:17] * toad_ ought to go eat...
[11:18] <cbreak> interesting information. Not only the kernel, also the firmware, the whole Posis Userland and everything else Apple has taken from BSD :)
[11:18] <toad_> hehe
[11:18] <toad_> sure but not the everyday stuff
[11:18] <toad_> GUI stuff
[11:18] <toad_> the finder etc
[11:18] <toad_> not to mention the apps
[11:20] <jay> toad_: inserts are timing out, but do you know which fnp message perhaps is involved, or anything more detailed?
[11:21] <toad_> jay: the timeouts on the insert states are way too short on many post queueing builds
[11:21] <jay> err any more details
[11:21] <toad_> because queueing slows everything down
[11:21] <toad_> that has been fixed recently
[11:21] * jay is at work and managing a group of students
[11:21] <jay> the FCP timeout?
[11:21] <toad_> no
[11:21] <toad_> at the fnp level
[11:21] <toad_> node/states/request/
[11:21] <jay> ok thought so
[11:22] <jay> i upped the fcp timeout to 10 minutes and it didn't help
[11:22] <toad_> InsertPending, TransferInsert, TransferInsertPending etc
[11:22] <toad_> jay: there are probably other bugs as well relating to timeouts
[11:23] <toad_> bbl
[11:23] <jay> the sim should expose these timeout errors i would think
[11:24] <toad_> eh?
[11:24] <jay> or i could run a few nodes here in the lab and monitor what happens with inserts
[11:24] <cbreak> the simulations are serial, only one request at a time, afaik.
[11:25] <jay> the timeouts are due to network conditions or logic errors?
[11:25] <cbreak> -> No Queueing, no Load -> not timeouts
[11:25] <toad_> the timeouts are due to queueing causing hops to take longer than expected
[11:25] <jay> im ahead of the sim then
[11:26] <jay> im trying to build a bootable debian w/ freenet and set up the lab with a test
[11:26] <jay> cd bootable i mean
[11:26] <toad_> cool
[11:27] <jay> set the nodes to debug and pipe the output over the network to a central machine
[11:27] <jay> my own completely unanonymous freenet
[11:28] <cbreak> and the DS? All in RAM?
[11:28] <jay> ill let you know when i get something even partially working
[11:28] <jay> theres disks on the machines
[11:28] <jay> and spare partitions
[11:28] <jay> they're all identical
[11:29] <jay> but they boot to XP and i can't really change that; it's easier to boot from the CD and mount a drive
[11:29] <jay> we use ghost to cast the image across the lan
[11:29] <cbreak> hmm... interesting :)
[11:30] <jay> i have to convince the faculty to allow me a small partition tho
[11:30] <jay> somehow
[11:30] * i2p_iip (~changate@host.teitel.net) has joined #freenet
[11:44] * hobx_ (~chatzilla@ankh.math.chalmers.se) has joined #freenet
[12:06] <jay> boy did that woman smell nice ;)
[12:06] <jay> ahh
[12:06] <jay> heh
[12:12] * hirvox (~hirvox@cs181027153.pp.htv.fi) Quit (Read error: 232 (Connection reset by peer))
[12:15] * d-ArkAngel (~Rob@213-131-104-86.onyx.net) has joined #freenet
[12:15] <d-ArkAngel> Hi all
[12:16] <cbreak> hi.
[12:16] <d-ArkAngel> is it my imagination or is the list seeming a little tense today? :-)
[12:18] <jay> hah
[12:18] <jay> this ain't nothing
[12:18] <jay> they're practically hugging each other in comparison to past debates
[12:19] <d-ArkAngel> lol, I've never seen such an argumentative discussion about an open source project :-)
[12:19] <jay> you haven't been around then
[12:19] <d-ArkAngel> still it proves everyone's "involved" with the spirit of the project, so I guess it's a good thing that everyone cares so much :-)
[12:19] <jay> no offense but just saying
[12:20] <d-ArkAngel> maybe the ones I've been involved with have been sissy :-)
[12:20] <jay> as long as the debate isn't personal it can be productive
[12:20] <d-ArkAngel> true as always.
[12:20] <jay> check out the linux kernel archives.. search for curse words and see theposts
[12:21] <jay> ive seen terrible arguments there :)
[12:21] <jay> personal insults etc.
[12:21] <jay> people tend to jump in around here when the argument gets inappropriate.. most people seem to respond well
[12:22] <jay> there's always the trolls of course but that's nothing new
[12:22] <d-ArkAngel> I've seen some of the tensions from kernel when it spilled over onto the motion lists about the removal of the Philips Webcam Drivers
[12:23] <jay> d-ArkAngel: how old are you, may i ask?
[12:24] <d-ArkAngel> 24
[12:24] * kers (~kers@h79n2fls33o847.telia.com) Quit (Read error: 110 (Connection timed out))
[12:24] <jay> you code in any langs?
[12:24] <d-ArkAngel> php, java
[12:24] <jay> ah cool
[12:24] <jay> id like to learn php myself
[12:24] <d-ArkAngel> it's very like perl, which I've also dabbled with
[12:24] <jay> ugh perl.. the language of chaos
[12:25] <jay> eheh i couldn't get into it
[12:25] <jay> u a sysadmin?
[12:25] <d-ArkAngel> yeah, php is like perl without the chaos :-)
[12:25] <jay> yup
[12:25] <jay> python is good too
[12:25] <jay> usually sysadmins know perl inside and out
[12:25] <d-ArkAngel> sysadmin... well I'm the "IT Director" for a Auction House, but since I'm currently the only IT person... :-)
[12:26] <jay> yer the everything man
[12:26] <d-ArkAngel> I also run a very large website, which I'm the only root on....
[12:26] <jay> btw im looking for a honda shadow .. how much ?
[12:26] <jay> heh
[12:27] <d-ArkAngel> well I might be able to get you a 1/45th scale model of one :-)
[12:27] <jay> lol
[12:27] <jay> nascar?
[12:27] <d-ArkAngel> We're a specelist toy auctioeers ;-) (technicaly the worlds largest)
[12:27] <jay> dale jr's #8 car
[12:27] <jay> ah cool
[12:28] <d-ArkAngel> I show you the website, but I'm ashamed to be associated with it :-) it's still part of the "old system" that I'm in the process of getting rid of.
[12:29] <d-ArkAngel> www.vectis.co.uk (what the hell, it's not like I care what people think of me :-) )
[12:30] <jay> u shouldn't
[12:30] <jay> :).. brb
[12:31] <d-ArkAngel> k
[12:40] * weilawei (fwuser@city.northampton.ma.us) has joined #freenet
[12:40] <weilawei> hewwo
[12:47] <jay> d-ArkAngel: site looks good
[12:49] <cbreak> site consists 25% of borders... :)
[12:50] * pittaman (~mistery_b@d51A410CA.kabel.telenet.be) has joined #freenet
[12:51] <jay> that's how sites layout content with just html
[12:51] <jay> err borders?
[12:51] <jay> u mean tables
[12:52] <jay> massive javascript too
[12:52] <cbreak> left and right the site is empty.
[12:53] <cbreak> somehow the whole content is centered, with massive, white borders on each side.
[12:54] <jay> html is ugly and shitty.. i don't spend too much time looking at it anymore
[12:54] * weilawei (fwuser@city.northampton.ma.us) Quit (Client Quit)
[12:54] <jay> leave that to the designers ;)
[13:00] <d-ArkAngel> yeah, it's one of the worst sites I've worked on :-) there is a new design in the pipline tho, so with luck the one I write to replace it should be half decent :-)
[13:00] <d-ArkAngel> anyway gents, anothe day over, and I'm off home.
[13:00] <d-ArkAngel> evening all.
[13:00] <cbreak> bye :)
[13:00] * d-ArkAngel (~Rob@213-131-104-86.onyx.net) Quit (Read error: 232 (Connection reset by peer))
[13:08] <jay> lates
[13:09] <toad_> hi
[13:09] * Ash-Fox (Hal-9000@aau121.neoplus.adsl.tpnet.pl) has joined #FreeNET
[13:10] <toad_> <jay> this ain't nothing
[13:10] <toad_> <jay> they're practically hugging each other in comparison to past debates
[13:11] <toad_> there's far more at stake on the tech list... most of the subscribers either haven't woken up yet, haven't checked their email, are in shock, or are waiting for The Gods to sort it out
[13:11] * eMBee (~mbaehr@sirius.iaeste.tuwien.ac.at) Quit (Read error: 110 (Connection timed out))
[13:15] <jay> friday is a slow list day, usually
[13:15] <jay> the avalanche should begin soon enough
[13:15] <jay> i hate that term, the gods
[13:16] <jay> newsbyte's
[13:16] <jay> the developers perhaps
[13:16] <toad_> newsbyte's
[13:16] <toad_> i object to it too
[13:16] <jay> im not sure u made it clear enough what this means
[13:16] <toad_> even though i just used it
[13:17] <jay> heh
[13:17] <toad_> capitalizing it suggests its use in a sarcastic manner
[13:17] <jay> but ian doesn't feel u can make conclusions just yet
[13:17] <toad_> i think
[13:17] <jay> well God and god are slightly different
[13:17] <jay> but that's OT ;)
[13:17] <toad_> :)
[13:18] <jay> good work on the sims tho man.. one way or another
[13:18] <jay> long time coming if u think about it
[13:18] <toad_> okay, i've killed one thread
[13:19] <toad_> what should i simulate now?
[13:19] <jay> ian should spell out what kind of sim you should now perform in order to give him the info he would like to see
[13:19] * toad_ thinks 75x25 would probably be a good idea
[13:20] <toad_> sqrt(N)*1.4
[13:20] <toad_> so 75 would be 12
[13:20] * guido^pe (~unknown@dsl-213-023-232-206.arcor-ip.net) has joined #freenet
[13:20] * toad_ simulates with 12, 10, and 15
[13:20] <toad_> hopefully this will run pretty fast
[13:23] * greycat (~wooledg@192.35.79.70) has joined #freenet
[13:23] <toad_> hi greycat
[13:24] <greycat> hi
[13:25] <toad_> running 10, 12, 15 @ 75x25...
[13:30] * i2p_iip (~changate@host.teitel.net) Quit (Remote closed the connection)
[13:30] * i2p_iip (~changate@host.teitel.net) has joined #freenet
[13:48] * Disconnected.
[23:33] * hapi (~PircBot@newton.ncc.edu) has joined #freenet
[23:33] * Topic is 'Upgrade to 5096 (5095 is mandatory) | Unstable: Upgrade to 60241 (60234 is mandatory) | Channel logs: http://newton.matcmp.ncc.edu/~lockej/freenet/chanlog/ | #freenet-politics and #freenet-chat are available for offtopic discussions'
[23:33] * Set by toad_ on Fri Sep 24 14:57:57 EDT 2004
These logs were automatically created by Jay Oliveri with his gimp hapi on irc.freenode.net.