Timestamps are in GMT/BST.
[0:10] * linyos (~asdfasdf@adsl-68-252-188-62.dsl.emhril.ameritech.net) has joined #freenet
[0:11] <linyos> i've been thinking a little more about my new network model
[0:11] <linyos> remember the grid computing thing - a way to access this pool of 'computons' that handle all your computing demands
[0:12] <linyos> well, the natural way to access that might be a big stupid opaque encrypted data stream
[0:12] <linyos> that shoots right in to the dark computon-net thing
[0:13] <linyos> what that means: you could have an excuse for shooting an opaque data stream through (say) the chinese internet or whatever
[0:13] <linyos> because it would be a totally legit thing that everybody does
[0:13] <linyos> and it just happens to be a perfect privacy primitive
[0:13] <linyos> because it betrays basically no information other than the fact that you have a link
[0:14] <linyos> (the destination ip would be meaningless)
[0:14] <linyos> so, you shoot your super stream right on through the firewall, into the grid computing center, and then you can compute and network just like anybody else
[0:15] <linyos> and nobody knows what the hell you're doing
[0:17] <linyos> anyway, that's not to mention the fact that such a system is inherently so much superior to the crappy internet
[0:18] <linyos> (eg: you fire up your 8mbit dsl link to the local datacenter where you can buy computing resources and network at fiber speeds)
[0:19] <linyos> you could do all your computing there and just receive a vector graphics stream over your link
[0:19] <linyos> whatever
[0:21] <linyos> toad_: re your complaint about anonymous cash, that is not a problem
[0:21] <linyos> people would buy computons under their real identity
[0:22] <linyos> if you are into anonymity, that would just mean that you use your computons to implement a freenet network over the dark grid
[0:22] <linyos> (however that works, it doesn't matter for the principle to stand)
[0:24] <linyos> btw, such a thing would be totally backwards compatible with IP routing: you could implement it with your computons in the darknet
[0:26] <linyos> so the network basically looks like: [alice] <--data stream--> [grid darknet where all sorts of stuff goes on rather privately] <-- data stream --> [bob]
[0:26] <linyos> outside the darknet grid, all the NSA sees are persistent encrypted data streams
[0:27] <linyos> inside the darknet, everything is privatized (from the processors to the disks to the links)
[0:27] <linyos> so there is little room for eavesdropping (unlike the internet where you just dump the traffic at the routers)
[0:28] <linyos> the big point is that this is a superior kind of network so you could move a ton of internet users onto it quickly. that makes it politically difficult to suppress.
[0:29] <linyos> it presses the question, you know: do you create the locked down TCPAnet of the future, or allow this thing to move internet activity into the shadows
[0:29] <linyos> i'm betting that the latter option could be taken
[0:32] <linyos> [9:36] <toad_> but IMHO it's unrealistic long term to expect large parts of the inet to be totally free
[0:32] <linyos> nah
[0:33] <linyos> as long as you have the legal right to own a rack of computers, sell their computing power, and connect them with wires to the other dude's rack of computers
[0:33] <linyos> you're basically fine
[0:34] <linyos> and frankly, that is just about as relaxed a jurisdictional assumption as you can get
[0:34] <linyos> if that's not allowed, you're left with the TCPAnet
[0:34] <linyos> (meaning the total locking down of all communications)
[0:35] <linyos> which is a plausible outcome but i sure don't hope for it
[0:36] <linyos> also "grid computing" has a big buzz around it. it doesn't seem outright subversive to the suits
[0:36] <linyos> (even though my little secret is that: it is precisely that)
[0:38] <linyos> btw, think of all the stuff you could do if you had gigabit links to everyone. there's a ton of wild applications.
[0:39] <linyos> not to mention the totally unexplored paradigm changes that distributed, piecemeal computing could have
[0:40] <linyos> it is radically different from the traditional computer... the non-locality of the activity
[0:45] * phweak (~phweak@dsl-62-3-71-168.zen.co.uk) Quit (Read error: 104 (Connection reset by peer))
[0:45] * phweak (~phweak@dsl-62-3-71-168.zen.co.uk) has joined #freenet
[0:46] <linyos> a concrete, conservative example of what might happen within the darkgrid might help. say you can buy a processor in the grid. you can link your processor to N of your friends' processors. then you can, for example, route messages or streams through them.
[0:46] <linyos> you can create an onion that depends on the integrity of N friends
[0:46] <linyos> and then small-world route the message out to anyone through the big web of friend-links
[0:46] <linyos> that's just one grid construction
[0:48] <linyos> and the only people who know a damn thing about what the hell you are doing in that grid are precisely those people you choose to entrust with that responsibility
[0:48] <linyos> (i suppose the existence of links might be somewhat more public, but that is hardly a big deal)
[0:49] <linyos> and there are good ways to break the association between the incoming-to-the-grid data stream and what goes on inside
[0:51] <linyos> but the fundamental victory here is that we've found out a way to get a single useful, opaque, private data stream through the internet without sounding alarms to the NSA
[0:52] <linyos> (or even bypass the internet entirely, if you can get a line directly to a grid access point, say by dsl)
[0:53] * rah (~rah@fennel.arb.bash.sh) Quit (Excess Flood)
[0:53] * rah (~rah@fennel.arb.bash.sh) has joined #freenet
[1:23] * MrNaughty (MrNaughty@d199-126-25-30.abhsia.telus.net) has joined #freenet
[1:23] * MrNaughty (MrNaughty@d199-126-25-30.abhsia.telus.net) Quit (Remote closed the connection)
[1:24] * MrNaughty (MrNaughty@d199-126-25-30.abhsia.telus.net) has joined #freenet
[2:14] * rah (~rah@fennel.arb.bash.sh) Quit (Read error: 60 (Operation timed out))
[2:24] * rah (~rah@fennel.arb.bash.sh) has joined #freenet
[2:35] * Rom|Sleep (Romster@wrnax2-178.dialup.optusnet.com.au) has joined #freenet
[2:40] * moskau23 (~Miranda@dsl-082-083-078-178.arcor-ip.net) Quit ("Miranda IM! Smaller, Faster, Easier. http://miranda-im.org")
[2:42] * moskau23 (~Miranda@dsl-082-083-078-178.arcor-ip.net) has joined #freenet
[2:45] * moskau23 (~Miranda@dsl-082-083-078-178.arcor-ip.net) Quit (Client Quit)
[2:48] * moskau23 (~Miranda@dsl-082-083-078-178.arcor-ip.net) has joined #freenet
[2:54] * gvdm (~gvdm@210-246-12-102.paradise.net.nz) has joined #freenet
[2:56] * moskau23 (~Miranda@dsl-082-083-078-178.arcor-ip.net) Quit ("Miranda IM! Smaller, Faster, Easier. http://miranda-im.org")
[2:59] * Romster (Romster@wrnax3-030.dialup.optusnet.com.au) Quit (Read error: 113 (No route to host))
[3:29] * mazzanet (~irc@mazzanet.user) has joined #freenet
[3:55] * m[a]zzanet (~irc@mazzanet.user) has joined #freenet
[3:58] * sanity (~ian@81-178-126-192.dsl.pipex.com) has joined #freenet
[4:08] * mazzanet (~irc@mazzanet.user) Quit (Read error: 110 (Connection timed out))
[4:08] * m[a]zzanet is now known as mazzanet
[4:27] * sanity (~ian@81-178-126-192.dsl.pipex.com) Quit (Read error: 110 (Connection timed out))
[4:43] * gvdm (~gvdm@210-246-12-102.paradise.net.nz) Quit ("I'm quitting")
[4:52] * gvdm (~gvdm@210-246-12-102.paradise.net.nz) has joined #freenet
[5:07] * sanity (~ian@81-178-126-192.dsl.pipex.com) has joined #freenet
[5:09] * gvdm (~gvdm@210-246-12-102.paradise.net.nz) Quit (Remote closed the connection)
[5:11] * rah (~rah@fennel.arb.bash.sh) Quit (Read error: 60 (Operation timed out))
[5:14] * gvdm (~gvdm@210-246-12-102.paradise.net.nz) has joined #freenet
[5:25] * MrNaught (MrNaughty@d199-126-25-30.abhsia.telus.net) has joined #freenet
[5:40] * moskau23 (~Miranda@dsl-082-082-235-059.arcor-ip.net) has joined #freenet
[6:03] * NullAcht15 (~NullAcht1@dsl-082-082-139-063.arcor-ip.net) has joined #freenet
[6:17] * nextgens (~nextgens@jabber.hst.ru) has left #freenet
[6:32] * fridim (~fridim@232-74-118-80.kaptech.net) has joined #freenet
[6:32] <fridim> yop yop
[6:58] * linyos (~asdfasdf@adsl-68-252-188-62.dsl.emhril.ameritech.net) has left #freenet
[7:09] * NullAcht15 (~NullAcht1@dsl-082-082-139-063.arcor-ip.net) Quit (Remote closed the connection)
[7:37] * Romster (Romster@wrnax3-007.dialup.optusnet.com.au) has joined #freenet
[7:47] <hobx_> gof gof
[7:49] * gvdm (~gvdm@210-246-12-102.paradise.net.nz) Quit (Read error: 110 (Connection timed out))
[8:01] * Rom|Sleep (Romster@wrnax2-178.dialup.optusnet.com.au) Quit (Read error: 113 (No route to host))
[8:39] * Romster (Romster@wrnax3-007.dialup.optusnet.com.au) Quit (Read error: 60 (Operation timed out))
[8:40] * fridim (~fridim@232-74-118-80.kaptech.net) Quit ("leaving")
[8:57] * nextgens (~nextgens@jabber.hst.ru) has joined #freenet
[8:58] <nextgens> hi
[8:58] <nextgens> toad_: here ?
[9:28] * Ash-Fox (UNKNOWN@edj136.neoplus.adsl.tpnet.pl) Quit (Nick collision from services.)
[9:28] * Ash-Fox (UNKNOWN@ede161.neoplus.adsl.tpnet.pl) has joined #FreeNET
[9:36] * FallingBuzzard (~srademach@66.151.22.70) has joined #freenet
[9:37] * MrNaughty (MrNaughty@d199-126-25-30.abhsia.telus.net) Quit (Read error: 54 (Connection reset by peer))
[9:37] * MrNaught (MrNaughty@d199-126-25-30.abhsia.telus.net) Quit (Read error: 104 (Connection reset by peer))
[9:37] * MrNaughty (MrNaughty@d199-126-25-30.abhsia.telus.net) has joined #freenet
[9:39] * MrNaughty (MrNaughty@d199-126-25-30.abhsia.telus.net) Quit (Client Quit)
[9:47] <toad_> nextgens: yeah
[9:48] <nextgens> can you start a /query please ?
[9:50] <toad_> huh?
[10:29] <hobx_> I need a qurery!
[10:29] <hobx_> Somebody set up an encrypted socket!
[10:33] <nextgens> lol
[10:33] <nextgens> hobx: I can't start one from here
[10:35] * nextgens is using a jabber gateway to access IRC :-P
[10:52] <hobx_> You just have to reroute the encryption through a recursive algorithm?
[11:10] * OnTheBaII (~me@host81-129-68-227.range81-129.btcentralplus.com) Quit (Remote closed the connection)
[11:16] * Hory (~Miranda@82.78.27.85) has joined #FreeNet
[11:34] * NullAcht15 (~NullAcht1@dsl-082-082-139-063.arcor-ip.net) has joined #freenet
[11:41] * hobx_ (~ag@ankh.math.chalmers.se) Quit (Remote closed the connection)
[11:51] * Romster (Romster@wrnax2-082.dialup.optusnet.com.au) has joined #freenet
[12:27] * moskau23 (~Miranda@dsl-082-082-235-059.arcor-ip.net) Quit ("Miranda IM! Smaller, Faster, Easier. http://miranda-im.org")
[12:46] * moskau23 (~Miranda@dsl-082-082-235-059.arcor-ip.net) has joined #freenet
[12:48] * moskau23 (~Miranda@dsl-082-082-235-059.arcor-ip.net) Quit (Client Quit)
[12:52] * moskau23 (~Miranda@dsl-082-082-235-059.arcor-ip.net) has joined #freenet
[12:52] * moskau23 (~Miranda@dsl-082-082-235-059.arcor-ip.net) Quit (Read error: 131 (Connection reset by peer))
[12:54] * moskau23 (~Miranda@dsl-082-082-235-059.arcor-ip.net) has joined #freenet
[13:21] * cbreak (~cbreak@84-72-98-199.dclient.hispeed.ch) has joined #freenet
[14:02] * rah (~rah@cpc1-nott1-3-0-cust206.nott.cable.ntl.com) has joined #freenet
[14:35] * goatee (~goatee@ip216-239-81-101.vif.net) Quit (Read error: 110 (Connection timed out))
[15:06] * moskau23 (~Miranda@dsl-082-082-235-059.arcor-ip.net) Quit ("Miranda IM! Smaller, Faster, Easier. http://miranda-im.org")
[15:22] * moskawa23 (~Miranda@dsl-082-082-235-059.arcor-ip.net) has joined #freenet
[15:26] * pittaman (~mistery_b@d5152629C.access.telenet.be) has joined #freenet
[15:38] * pittaman (~mistery_b@d5152629C.access.telenet.be) Quit (Client Quit)
[16:10] * Hory (~Miranda@82.78.27.85) Quit ("CyberLore.net - Recommendations on the best games, freeware and websites.")
[16:17] * nextgens (~nextgens@jabber.hst.ru) has left #freenet
[16:21] * moskawa23 (~Miranda@dsl-082-082-235-059.arcor-ip.net) Quit ("Miranda IM! Smaller, Faster, Easier. http://miranda-im.org")
[16:28] * Romster1 (Romster@wrnax3-243.dialup.optusnet.com.au) has joined #freenet
[16:38] * moskau23 (~Miranda@dsl-082-082-235-059.arcor-ip.net) has joined #freenet
[16:39] * Romster (Romster@wrnax2-082.dialup.optusnet.com.au) Quit (Read error: 60 (Operation timed out))
[17:27] * moskau23 (~Miranda@dsl-082-082-235-059.arcor-ip.net) Quit ("Miranda IM! Smaller, Faster, Easier. http://miranda-im.org")
[17:30] * moskau23 (~Miranda@dsl-082-082-235-059.arcor-ip.net) has joined #freenet
[17:31] * sleon|tuX (test@e180022090.adsl.alicedsl.de) has joined #freenet
[17:37] * moskau23 (~Miranda@dsl-082-082-235-059.arcor-ip.net) Quit ("Miranda IM! Smaller, Faster, Easier. http://miranda-im.org")
[17:42] * sleon|tuX (test@e180022090.adsl.alicedsl.de) Quit ("Leaving")
[17:56] <toad_> if i buy an audiobook from the US, how long will it be held up by customs? I'm in the UK
[17:59] * NullAcht15 (~NullAcht1@dsl-082-082-139-063.arcor-ip.net) Quit (Remote closed the connection)
[18:06] * moskawa23 (~Miranda@dsl-082-082-235-059.arcor-ip.net) has joined #freenet
[18:09] <sanity> toad: we don't cache locally made requests: correct?
[18:10] <sanity> toad: if not, why would a splitfile take less time to request the second time around?
[18:10] <toad_> sanity: we cache them the same way we cache anything else
[18:10] <toad_> otherwise, it would be a dead giveaway
[18:11] <toad_> it's pretty bad either way though
[18:11] <toad_> we need premix routing
[18:11] * moskawa23 (~Miranda@dsl-082-082-235-059.arcor-ip.net) Quit ("Miranda IM! Smaller, Faster, Easier. http://miranda-im.org")
[18:11] <sanity> what about probabilistic caching?
[18:11] <toad_> and IMHO premix routing with trusted links requires us to expose the network topology
[18:11] <toad_> right
[18:11] <toad_> we cache locally requested content with the same probability as remotely requested content
[18:11] <toad_> which is 1.0 if the store is less than 90% full
[18:11] <toad_> and in practice 1.0 if the store is less than 100% full
[18:11] <sanity> but won't locally requested content have a higher htl?
[18:12] <toad_> on average yes, but we don't take that into account directly
[18:12] <sanity> also, why wouldn't people's datastores then get flushed every time they request a large file?
[18:12] <toad_> it will have a higher hopsSinceReset
[18:12] <toad_> meaning that it will have a lower probability of being cached
[18:12] <toad_> people's stores do
[18:12] <toad_> we have to download the stuff into the store
[18:12] <toad_> even if we then purge it
[18:13] <toad_> you need to have store much bigger than downloads
[18:13] <toad_> unfortunately this is unusual
[18:14] <sanity> so doesn't this make it pretty trivial to determine whether someone has themselves requested a file? just request if from them and see how quickly they retrieve it?
[18:14] <toad_> somebody else might have requested it from them
[18:14] <sanity> someone else is unlikely to request an entire file from any one person
[18:14] <toad_> the only thing we can do about it is premix routing
[18:15] <toad_> we CANNOT make the node not cache a file fetched by the user
[18:15] <sanity> why not?
[18:15] <toad_> if we do, then an attacker can know for certain that the user requested the file
[18:15] <sanity> how?
[18:15] <toad_> user requests bunch of files from attacker node
[18:15] <toad_> attacker node returns them
[18:15] <toad_> attacker node requests them from user at htl 0
[18:15] <toad_> attacker node sees that NONE of them are on the user node
[18:16] <toad_> attacker node puts 2 and 2 together
[18:16] <toad_> as long as there are enough chunks
[18:16] <toad_> if it's only one, he gains less info - but freesites consist of many linked files
[18:16] <toad_> not just one file
[18:18] <sanity> what makes that any worse than the other attack: attacker gets request from user for file, attacker requests entire file from user and sees how long it takes.
[18:18] <sanity> and that won't work if we get rid of HTL
[18:19] <toad_> hmm?
[18:19] <toad_> getting rid of HTL isn't on the cards IIRC
[18:20] <toad_> sanity: if the attacker has got a request from the user for the file, then he knows there's a good chance the user cached it
[18:20] <sanity> so right now a remote user can tell what files i have personally requested
[18:20] <toad_> unless we make the user node always drop locally requested files
[18:20] <toad_> in which case the attacker can be far more certain that they were requested locally
[18:22] * FallingBuzzard (~srademach@66.151.22.70) has left #freenet
[18:22] <toad_> well that's always been my view on why we don't do it anyway
[18:23] <toad_> basically everything sucks until we get premix routing. we can improve matters slightly by dropping HTL, but the compromize we came up with is okay (max htl of 11, at 11 there is a 10% chance of dropping down to 10, at 1 there is a 20% chance of dropping down to 0)
[18:23] <sanity> so lets get this straight: right now, a remote user can tell what files i have personally requested from freenet?
[18:23] <toad_> where you can't actually send an htl 0 request
[18:23] <toad_> you have to send an htl 1 request
[18:23] <toad_> sanity: you have plausible deniability, and they are pcached just like every other file
[18:24] <toad_> sanity: but possibly. and having them deleted when you fetch them is not going to solve the problem either.
[18:25] <toad_> not having a reliable htl 1 does help with some of these attacks
[18:25] <toad_> but you _can_ time it
[18:27] <toad_> timing is significant - the longer it has been since you fetched the stuff, the less likely it is to be in your store
[18:27] <toad_> size of splitfiles is also important
[18:28] <toad_> if they are big you have more data to correlate
[18:29] <sanity> i think we should have something in the FAQ about this.
[18:31] <toad_> quite possibly, though this seems a change of heart relative to your usual optimism :)
[18:33] <sanity> i have never been happy about this issue
[18:33] <toad_> lets see
[18:34] <toad_> what exactly is the attack vector...
[18:34] <toad_> if you have a huge, non-full store, then an attacker can simply probe your store
[18:34] <toad_> and identify what has been requested through your node
[18:34] <toad_> they can then get a statistical estimate of which splitfiles you yourself have fetched
[18:34] <toad_> and sites too maybe
[18:34] <toad_> although they can't prove it
[18:35] <toad_> since it's quite possible some degenerate node has routed a load of stuff only to your node
[18:35] <toad_> on a network that works well they can be pretty confident though
[18:36] <toad_> if you have a full store, and the attacker has a connected node, they can see what you request and do correlation attacks... HTL helps slightly.
[18:36] <toad_> if we delete files after they are fetched locally, then they have to have a connected node to be sure.. but they can be almost absolutely sure if the splitfile is big enough.
[18:36] <toad_> so whichever way it goes, we have a problem
[18:37] <sanity> we could simply not allow splitfiles
[18:37] <toad_> no, we couldn't
[18:37] <Elly> toad_, remember
[18:37] <sanity> whhy not?
[18:37] <toad_> we could recommend that people not use splitfiles
[18:37] <toad_> we can't prevent splitfiles
[18:37] <sanity> that is true
[18:37] <Elly> "What's the simplest thing that could possibly work?"
[18:38] <toad_> also if we want fixed size keys we will need small splitfiles just for freesites
[18:38] <toad_> the vulnerability is easier if we have splitfiles, but you may still be able to gather some info from freesite browsing and/or frost
[18:39] <toad_> also IMHO splitfiles are of some value... and significant to a large part of our userbase
[18:39] <toad_> certainly without them we won't get much datastore stress until the network is enormous
[18:39] <toad_> there simply isn't that much text out there
[18:40] <Elly> also, a lot of useful shit just isn't plain text
[18:40] <Elly> heh
[18:40] <toad_> right
[18:40] <toad_> the diebold files are quite large
[18:40] <toad_> and i'd like to be able to use freenet for free no-advert largish-file hosting
[18:40] <toad_> for e.g. game mod sites
[18:41] <toad_> sanity: what do you think about exposing the network topology on a fixed links network?
[18:41] <Elly> won't the network topology be fairly obvious to anyone watching anyway?
[18:41] <Elly> I mean, freenet traffic is identifiable
[18:42] <sanity> toad: how do you find out about nodes you aren't directly connected to if you expose "near by" network topology?
[18:42] <toad_> hmmm?
[18:44] <toad_> sanity: the idea would be that nodes would tell you
[18:44] <sanity> i am not convinced by the attack on nodes which don't cache local requests
[18:45] <toad_> sanity: I connect to Ian, Oskar, and Thelema
[18:45] <toad_> Thelema connects to Fish and KenMan
[18:45] * goatee (~goatee@ip216-239-81-58.vif.net) has joined #freenet
[18:45] <sanity> yes, i suppose if you trust ian, oskar, and thelema to tell the truth then that is fair enough
[18:45] <toad_> KenMan connects to Paul and Ellie
[18:45] <toad_> well..
[18:45] <sanity> we will need to premix in 0.7 for sure
[18:45] <toad_> it's partly that
[18:45] <toad_> but also, you can corroborate them
[18:46] <sanity> how?
[18:46] <toad_> if you're going to set up a premix connection, you need to know you'll have a good anonymity set
[18:46] <toad_> if you know the graph out to 4 or 5 hops, you can identify it
[18:46] <toad_> the main attack is for a node to lie and invent an entire subnetwork
[18:46] <toad_> but if this is true, then it will not be connected to any of the rest of the network
[18:47] <sanity> true
[18:47] <toad_> it should be possible to identify subnetworks that are likely to be entirely fictional
[18:47] <toad_> as long as links are signed by both ends
[18:47] <toad_> obviously we would only expose the public keys, not the connection info, and the pubkeys would be different to the ones used to connect to the nodes
[18:47] <hobx> I don't tell the truth
[18:47] <hobx> Every statement I make, including this one, is a lie!
[18:47] <toad_> i'm not sure exactly how the math would work though
[18:47] <toad_> hobx: :)
[18:48] <hobx> I never drink either...
[18:48] <toad_> and you never fantasize about sexually harassing students
[18:48] <toad_> in public
[18:49] <toad_> sanity: well that's the only way i can figure to do premix routing on a fixed-links mesh
[18:49] <toad_> also it'd let us get a fairly accurate guess of the size of the network
[18:49] <sanity> toad: so what are you working on right now?
[18:49] <toad_> which might be nice
[18:49] <toad_> reading email
[18:50] <toad_> then i'm going to check out my simulations
[18:50] <toad_> i have a bunch of ideas on how to improve them
[18:50] <sanity> ok, i'm just worried that it doesn't appear that much progress is being made :-/
[18:50] <toad_> including a new gravity estimator that actually simulates gravity
[18:50] <toad_> yeah, that's my fault
[18:50] <sanity> gravity?!
[18:50] <toad_> the first NGR estimator used an approximation to gravity
[18:51] <sanity> ah, yes
[18:51] <toad_> it's just that the approximation really sucked
[18:51] <sanity> but why?
[18:51] <toad_> i wonder if a more accurate approximation might not do exactly what we want
[18:51] <sanity> and you can empirically test it against what we have now?
[18:51] * toad_ draws a graph
[18:51] <toad_> sanity: right
[18:51] <toad_> there's a bag of stuff i can test or tweak
[18:52] <Elly> toad_: Have nodes invent fake nodes with humorous addresses!
[18:52] <Elly> * Connected to: freenet.nsa.gov
[18:52] <Elly> =D
[18:52] <toad_> if none of it pans out we'll need a new algorithm, or we'll need to figure out a way to use the network with sucky psuccess's
[18:52] <sanity> which network?
[18:52] <toad_> Elly: no names. no addresses. no identifying info whatsoever, except a pubkey (which isn't the identity PK used for connecting), and a list of links, which are signed both ways
[18:52] <sanity> are you still referring to the fixed-links stuff?
[18:52] <toad_> yeah
[18:53] <toad_> actually NGR+LRU sucks in simulation too but that's another story :)
[18:53] <Elly> toad_: I vote we root nsa.gov and turn it into a node
[18:53] <toad_> well i haven't tried it with the gradual growth model actually, it might work well with that
[18:54] <toad_> Elly: sounds like a plan. seeya in jail and/or hell (from heaven).
[18:54] <toad_> :)
[18:54] <Elly> mmm, hell
[18:54] <Elly> the only problem is that they may want the box back later
[18:54] <Elly> hmmm
[18:55] <toad_> http://amphibian.dyndns.org/multi-whacky-graph.png
[18:55] <Elly> I don't even understand that graph
[18:55] <toad_> the best algorithm so far falls to around 85% psuccess with 9000 nodes
[18:55] <toad_> oddly enough none of them has actually gone over 10,000 nodes
[18:55] <sanity> toad: i just worry that we could spend potentially infinite amounts of time mucking around with simulations without making progress on 0.7 itself
[18:56] <toad_> it should recover a bit after, if they just need more requests
[18:56] <toad_> sanity: that is a possible problem. as is building a 0.7 which we have no idea whether it will work and spending another 7 years trying random things to make it work better.
[18:56] <sanity> ideally we need to find an algorithm that can route in a fixed links network, and having found it, implement it
[18:56] <toad_> absolutely
[18:56] <toad_> well, we have an algorithm that can route in a fixed links network
[18:57] <toad_> it's a question of how well do you need it to route
[18:57] <sanity> well, how well is it routing right now?
[18:57] <toad_> not well enough to scale to a decent number of nodes
[18:57] <toad_> but we don't really know because of all the load problems
[18:57] <toad_> although queueing has helped a good deal
[18:57] <toad_> hopefully that will be different in the new network anyway
[18:58] <toad_> (but i do
[18:58] <toad_> (but i do think we'll end up building two 0.7's...)
[18:58] <toad_> although we could just keep 0.5 as the non-fixed-links network
[18:58] <sanity> load problems?
[18:58] <sanity> you have load problems in the simulations?
[18:58] <toad_> no
[18:58] <toad_> in real life
[18:58] <sanity> well why are you bringing that into this conversation?
[18:58] <toad_> in the simulations we get about 85% psuccess at 9,000 nodes with the best algorithm
[18:59] <toad_> it will be much worse IRL
[18:59] <sanity> when data isn't found - why isn't it being found?
[18:59] <sanity> is it because it isn't in the network, or because it is in the network but can't be found?
[19:00] <toad_> http://amphibian.dyndns.org/multi-whacky-pathlength-graph.png
[19:00] <toad_> here's the path length
[19:00] <sanity> a path length less than 1?
[19:00] <sanity> surely that is psuccess?
[19:01] <sanity> what are the axis of that graph?
[19:02] <Elly> Welcome to a world where the air I breathe is mine, nothing to overwhelm me and nothing to cloud my mind...
[19:02] <toad_> http://amphibian.dyndns.org/multi-whacky-pathlength-graph.png
[19:02] <toad_> y axis is number of hops, goes up to 16
[19:02] <toad_> http://amphibian.dyndns.org/multi-whacky-graph.png is probability
[19:03] <sanity> ah, ok
[19:03] <toad_> Nulls: 0, Valid: 21916, average access count: 0.5013688629311919, max access cou
[19:03] <toad_> nt: 6, min access count: 0, fetchable: 21891, duplication max: 61, duplication m
[19:03] <toad_> in: 0, duplication mean: 15.994752692097098
[19:03] <toad_> in other words, 99%+ of the keys are fetchable
[19:03] <toad_> i.e. they exist in the network somewhere
[19:03] <toad_> this is with 1 insert to 1 request i.e. absurdly high datastore churn (this is beneficial)
[19:04] <toad_> Elly: you own the air?
[19:04] <sanity> why is the pathlength increasing with time?
[19:04] <sanity> are those networks growing?
[19:04] <toad_> path length increases with number of nodes
[19:04] <toad_> which is the x axis
[19:05] <toad_> not actually time
[19:05] <toad_> and yes it is growing
[19:05] <hobx> time is my friend
[19:05] <toad_> most of those look reasonably loggish
[19:05] <Elly> toad_: Damn right
[19:05] <Elly> no breathing my air!
[19:05] <toad_> and remarkably low
[19:05] <hobx> SPACE IS MY ENEMY!
[19:06] <toad_> Elly: /me regrets he cannot comply for more than a few minutes
[19:06] <toad_> under 10 hops for 9000 nodes on the best one
[19:07] <Elly> STOP BREATHING MY AIR, YOU COPYRIGHT/PATENT-INFRINGING SCUMBAGS
[19:07] <toad_> hmmm
[19:07] <toad_> slightly over 10 hops
[19:08] * IRCMonkey_ (~chatzilla@chello080108161048.5.12.vie.surfer.at) has joined #freenet
[19:08] <IRCMonkey_> hi
[19:08] <toad_> sanity: do you think it would affect it much if we batched estimator updates? it would be a problem in bootstrapping, but maybe not later on
[19:08] <toad_> it would let us route much quicker
[19:09] * toad_ avoids risque conversations with Elly
[19:09] * Elly doesn't
[19:10] <toad_> sanity: the psuccess doesn't seem to be directly connected to the path length, that's the odd thing
[19:10] <toad_> sanity: it seems to fall fairly consistently
[19:10] <Elly> maybe it is and your sim is just wrong?
[19:10] <toad_> whereas we can be reasonably sure that 20 hops will cover a huge network as far as path length goes...
[19:10] <toad_> Elly: maybe what is?
[19:12] <Elly> directly connected to the path length
[19:12] <toad_> sanity: what's a good target for 10,000 nodes in simulation?
[19:12] <Elly> it _should_ be
[19:12] <Elly> toad_: 102% success rate.
[19:12] <Elly> And it should cure cancer.
[19:12] <toad_> :)
[19:12] <Elly> And it should be precognizant.
[19:12] <toad_> well the decline doesn't seem to decelerate fast enough
[19:12] <toad_> if it was directly connected to the path length it would slow down
[19:13] <toad_> like the path length's increase does
[19:13] <toad_> brb
[19:15] <sanity> what is the pattern of requests to inserts?
[19:18] <toad_> on all but one line, we do an insert, add it to the active set and kick the LRU out of the active set, then we request a random file from the active set
[19:19] <toad_> i tried 10 requests to one insert, this sucked
[19:19] <toad_> i don't think it matters all that much as long lived content will probably be popular content
[19:20] <toad_> email me with more if you like; i'm going to get lunch
[19:20] <toad_> errr dinner
[19:20] <toad_> tea
[19:20] <toad_> :)
[19:20] <toad_> sanity: please do consider the question of what would be a tolerable psuccess target for 10k nodes...
[19:20] <toad_> bbiab
[19:21] <toad_> and whether we can batch updates (i suppose i'll just have to try it...)
[19:23] <sanity> toad: well, consider the effect of FEC
[19:23] * IRCMonkey_ (~chatzilla@chello080108161048.5.12.vie.surfer.at) Quit ("ChatZilla 0.9.61 [Mozilla rv:1.7.7/20050414]")
[19:24] <toad_> bbiab, email me
[19:25] <toad_> sanity: only applies to splitfiles... not manifests and DBRs
[19:25] <toad_> but maybe that's okay
[19:25] <toad_> due to popularity
[19:25] <toad_> bbiab
[19:25] * toad_ (toad@toad-with-underline.active.supporter.pdpc) Quit (Remote closed the connection)
[19:30] * hobx feels lonely and ignored.
[19:38] <Elly> Why, I wonder
[19:39] <Elly> is the entertainment industry
[19:39] <Elly> so STUPID?
[20:17] * cbreak (~cbreak@84-72-98-199.dclient.hispeed.ch) Quit ("leaving")
[20:53] * linyos (~asdfasdf@adsl-68-252-188-62.dsl.emhril.ameritech.net) has joined #freenet
[21:00] * toad_ (toad@toad-with-underline.active.supporter.pdpc) has joined #freenet
[21:00] * ChanServ sets mode +o toad_
[21:30] * Romster (Romster@wrnax1-045.dialup.optusnet.com.au) has joined #freenet
[21:34] <linyos> i wrote up some of my recent ideas in cute little one-page note
[21:36] <linyos> critics invited
[21:37] * Romster1 (Romster@wrnax3-243.dialup.optusnet.com.au) Quit (Read error: 148 (No route to host))
[22:00] * goatee (~goatee@ip216-239-81-58.vif.net) Quit (Read error: 110 (Connection timed out))
[22:47] <toad_> linyos: telling me that doesn't help me to criticise it
[22:47] <toad_> since i don't know where it is
[22:53] <linyos> on the way
[22:55] <toad_> linyos: didn't work
[22:55] <toad_> it said connection failure - timed out
[22:55] <linyos> hmm
[22:55] <linyos> firewall
[22:55] <toad_> your end
[22:55] <linyos> yeah, hang on a bit
[23:00] <linyos> i'll just start my web server
[23:00] <toad_> same problem
[23:02] <linyos> toad_: http://znex.org/alpha.pdf
[23:04] <linyos> it is not exactly newton's principia, but i was happy with my reasoning
[23:05] <toad_> what is it?
[23:06] <linyos> a summary of my thoughts about the whole they'll-detect-your-node-and-kill-you problem with freenet
[23:06] <toad_> oh
[23:06] <toad_> i remember
[23:06] <toad_> you're newsbyte
[23:06] * linyos is now known as mj1
[23:06] <mj1> who stole my nick!!
[23:06] * mj1 is now known as mjr_
[23:07] <mjr_> you leave for a couple years and somebody just goes and steals your nick...
[23:07] <mjr_> anyway, hi
[23:08] <toad_> the point about grid computing is dubious. the guy who tried it folded when people realised that they really couldn't make major dosh out of their internet connections. :)
[23:09] <mjr_> toad_: it's just a way to push the networking off the public, monitored wires and into a more private domain
[23:09] <mjr_> but the principles apply no matter what the alpha-minimizing technology is
[23:09] <toad_> mojonation, that's what it was called
[23:10] <toad_> so what's the conclusion? that there's no point trying to build a globally scalable darknet because you can't hide it? or that there's every point because sooner or later there will be a convenient service to hide it behind?
[23:11] <mjr_> it's a way to reason about the problem of eavesdropping
[23:11] <mjr_> either you fool eve into thinking your behavior is "normal"
[23:11] <mjr_> which is almost impossible because "typical web use" is so damn predictable
[23:12] <mjr_> or you change the normative use to reveal virtually no information to eve
[23:12] <mjr_> by surreptitiously making that a consequence of the latest new buzzword technology
[23:13] <toad_> heh
[23:13] <toad_> the latest new buzzword technology will still not be CBR
[23:13] <toad_> it will have patterns
[23:13] <toad_> unless people start leaving VoIP channels on 24x7
[23:13] <toad_> to more than one person
[23:13] <toad_> which is almost conceivable but not quite
[23:13] <toad_> s/VoIP/H.323
[23:13] <toad_> videophone
[23:14] <mjr_> as i imagine it (to minimize alpha) you would only use the internet to connect to the private grid at one access point
[23:14] <mjr_> that would be a persistent (cbr if you want) connection
[23:15] <toad_> well it wouldn't be CBR
[23:15] <mjr_> eve could try to get into the private grid, by participating in various ways, but that is a way bigger challenge
[23:15] <toad_> you would want to optimize bandwidth use in order to get best performance if you have several connections
[23:15] <mjr_> and one you can work around
[23:15] <toad_> therefore no CBR
[23:15] <toad_> unless paranoid privacy is a key goal
[23:15] <mjr_> toad_: there would only be one connection. you'd do all your networking on the dark grid.
[23:15] <toad_> which it won't be to that extent
[23:16] <mjr_> and the grid is privatized so there is no router to wiretap
[23:16] <toad_> mjr_: in order to have a grid you must have more than one connection
[23:16] <mjr_> the connections in the grid would be point-to-point
[23:16] * goatee (~goatee@ip216-239-81-189.vif.net) has joined #freenet
[23:16] <mjr_> like, say i rented a computer on this hypothetical grid
[23:16] <toad_> mjr_: and being able to sell bandwidth is not sufficient reason to have to deal with substantially increased latency, because bandwidth is not a realistic commodity. CPU power might be, but not bandwidth.
[23:17] <mjr_> and i ran an ethernet cable from my computer in the data center to yours and oskar's
[23:17] <mjr_> then we can network all we want at gigabit speeds
[23:17] <toad_> and we absolutely must not tie freenet to anonymous digicash!
[23:17] <mjr_> expand that graph a little bit, and you get a small-world mesh
[23:17] <toad_> mjr_: you mean user-owned physical hardware distinct from the internet as we know it?
[23:17] <toad_> _that_ would be cool
[23:17] <mjr_> toad_: the latency would be less than that of the existing internet. you could send packets instantly at fiber speeds to the other nodes you are plugged in to
[23:17] <toad_> however, if eve knows one node, she can trace pretty much the whole network with a little effort if they are physical connections
[23:18] <mjr_> she can only trace the physical links, which does not betray any information about how you are using those encrypted gigabit links
[23:18] * toad_ is affiliated to a local co-op that is attempting to build a wireless grid
[23:18] <toad_> mjr_: you can only get a gigabit per second over a link which is at least a gigabit per second
[23:19] <mjr_> the point is that once you are off the internet, where every packet is intercepted and examined, you have a lot of a room for privacy
[23:19] <toad_> mjr_: no amount of jiggery pokery, and no amount of capitalism, can squeeze more out
[23:19] <toad_> you cannot tunnel a gigabit over a 57.6kbps connection
[23:19] <mjr_> toad_: the grid processors would be physically congregated in various datacenters. the centers could have huge direct fiber links to each other
[23:20] <toad_> in which case They would tap _EVERYTHING_ at the datacenters
[23:20] <toad_> makes it even easier for them than it is now
[23:21] <mjr_> how do you tap an encrypted, CBR ethernet link?
[23:21] <mjr_> but, yeah, you could make private networks illegal
[23:21] <toad_> it's a central point of failure
[23:21] <toad_> and it's not really a private network
[23:21] <toad_> it's a huge great megacorp hosted data center
[23:21] <toad_> which WILL be interfered with by a hostile government
[23:21] <toad_> and the megacorps will be only too happy to play ball
[23:22] <toad_> because they provide the censorship hardware the government uses on the open net
[23:22] <toad_> sounds like a silly badly thought out libertarian wet dream to me :)
[23:22] <mjr_> well, it's that, or they arrest you for running your freenet node, which you can never (as my pdf note explains) sufficiently hide against a background of web traffic
[23:23] <mjr_> i mean, is that so much better?
[23:23] <toad_> neither option is acceptable
[23:23] <toad_> the datacenters thing is highly implausible though
[23:23] <toad_> it relies on massive central failure points
[23:23] <toad_> and it relies on them being used by Ordinary People, which they won't be
[23:23] <toad_> because Ordinary People have no reason to use them
[23:23] <mjr_> the hope is that private datacenters (which anyone could run, at any scale) would be permitted without horrible government intervention (like mandatory logging of all activity) in most liberal democracies
[23:24] <toad_> the heaviest thing they use is a game. and a game will have to run locally, because it needs a massive amount of bandwidth to the screen.
[23:24] <mjr_> nah, that's just my point. you could have a technology that's way better than what people get today with the old, bad, slow, latent internet
[23:24] <toad_> mjr_: there would be economies of scale - you want low latency to the rest of the world, so you go with the biggest provider
[23:25] <toad_> end result, one or two megacorps own 99% of the market
[23:25] <toad_> and it's all centralized in a few very large datacenters
[23:25] <mjr_> regulation is a threat, i agree
[23:25] <mjr_> but why not just lock down the internet then?
[23:25] <mjr_> with TCPA
[23:25] <toad_> you should meet oskar, you think the same way, you're just slightly less jadedly cynical :)
[23:25] <toad_> you can do that
[23:26] <toad_> I happen to think that guerilla networking will continue long after that happens, if it does eventually happen
[23:26] <mjr_> i mean, you have to draw the line somewhere and hope that freedom is allowed somewhere
[23:26] <toad_> even if you have to get black market hardware to do it
[23:26] <toad_> sure
[23:26] <mjr_> and i'm quite optimistic that in the west people will be able to network their computers without odious government intervention
[23:27] <toad_> but your proposals rely on naive assumptions. anyone who's ever read the mailing lists will tell you that big megacorporate central failure points are bad.
[23:27] <toad_> and the market for consumer remote computing really doesn't exist
[23:27] <toad_> and may never exist
[23:28] <mjr_> you could share files at instantaneous speeds, store all the data you want on secure disks, do big computations instantly, and things i haven't thought of yet
[23:28] <toad_> and you'd need a low latency connection with a fair bit of bandwidth to your end in order to use it anyway
[23:28] <toad_> no you couldn't
[23:28] <toad_> you could transfer files within the cluster at very high speeds
[23:28] <toad_> if you wanted to actually download them to your PC to use them, that would take ages
[23:28] <toad_> you could have good latency within the cluster
[23:29] <toad_> but you would still have to deal with the inevitable latency from your PC to the cluster
[23:29] <toad_> if you want to actually use it
[23:29] <toad_> and "secure disks" is highly dubious
[23:30] <toad_> why is my data more secure if it runs on a Microsoft hosted server? For a fee they may provide a decent backup solution, but in terms of who has access, it's less secure
[23:30] <mjr_> so you either use it via some kind of dsl-speed vector graphics stream, which would work great for most stuff, or you bite the bullet and revert to doing the computation locally where you can get the result instantly
[23:30] <mjr_> your data could be secure on oskar's server because you trust oskar not to let the NSA access his disks, whatever
[23:30] <toad_> sure
[23:30] <toad_> but in practice you wouldn't use oskar's server
[23:30] <toad_> you'd use microsoft's
[23:31] <toad_> because everyone else does, and you have to be on the same cluster to get scarily fast transfers
[23:31] <mjr_> there is certainly room for thought about blue-sky ideas like this
[23:31] <toad_> i still say the market for remote computing simply does not exist at a consumer level and probably won't for a very very long time
[23:32] <toad_> i think consumer/collectively owned network hardware is more likely to see widespread adoption (and that's still a long shot)
[23:33] <mjr_> maybe, maybe not. the nature of the new technology i described is sort of secondary to the main issue, which is eve and how to deal with her
[23:33] <toad_> well, i need to go to bed
[23:33] <mjr_> good night
[23:33] <toad_> i suggest you post any important ideas to tech@freenetproject.org
[23:34] <mjr_> nah, even here is too public
[23:34] <toad_> well then send me an encrypted email ! :)
[23:34] <toad_> if my box has been rooted by the NSA then so has yours :)
[23:35] <toad_> well probably not but only because they don't consider you to be a big player :)
[23:35] <mjr_> yeah, i hate how they keep rooting my boxes
[23:35] <toad_> lol
[23:35] <toad_> bbl zzzzzzzzzzzzzzzzzzzzzzzzzzz
[23:35] <mjr_> later
[23:52] * mjr_ works on his paper
These logs were automatically created by Jay Oliveri with his gimp hapi on irc.freenode.net.
