Timestamps are in GMT/BST.
[0:50] * mjr_ works more on his paper
[0:51] * Romster (Romster@wrnax1-045.dialup.optusnet.com.au) Quit (Read error: 113 (No route to host))
[1:28] * Romster (Romster@203.129.151.119) has joined #freenet
[1:37] * Romster1 (Romster@203.129.151.119) has joined #freenet
[1:57] * Romster (Romster@203.129.151.119) Quit (Read error: 110 (Connection timed out))
[2:12] * mjr_ thinks toad is going to love the libertopian turn this thing has taken
[2:17] * MrNaughty (MrNaughty@d199-126-25-30.abhsia.telus.net) has joined #freenet
[2:18] * Romster1 is now known as Romster
[2:18] * Romster (Romster@203.129.151.119) Quit (Nick collision from services.)
[2:18] * Romster (Romster@203.129.151.119) has joined #freenet
[2:45] * moskau23 (~Miranda@dsl-082-082-235-059.arcor-ip.net) has joined #freenet
[3:05] * moskau23 (~Miranda@dsl-082-082-235-059.arcor-ip.net) Quit ("Miranda IM! Smaller, Faster, Easier. http://miranda-im.org")
[3:08] * moskau23 (~Miranda@dsl-082-082-235-059.arcor-ip.net) has joined #freenet
[3:10] * moskau23 (~Miranda@dsl-082-082-235-059.arcor-ip.net) Quit (Client Quit)
[3:36] * gvdm (~gvdm@unibus.interface.org.nz) has joined #freenet
[3:48] * gvdm_ (~gvdm@unibus.interface.org.nz) has joined #freenet
[3:48] * gvdm_ (~gvdm@unibus.interface.org.nz) Quit (Client Quit)
[4:32] * gvdm (~gvdm@unibus.interface.org.nz) Quit (Read error: 104 (Connection reset by peer))
[4:43] * gvdm (~gvdm@unibus.interface.org.nz) has joined #freenet
[5:06] * gvdm (~gvdm@unibus.interface.org.nz) Quit (Read error: 60 (Operation timed out))
[5:11] * mjr_ revises his logic
[5:19] <mjr_> i am always right, but the reason why i am right shifts around a little.
[5:44] <mjr_> it bothers me that a person who thinks can take all day to write a few paragraphs, while unthinking jackasses write volumes in no time
[5:44] <mjr_> the ironies of this world, d00d
[5:48] * gvdm (~gvdm@210-246-18-159.paradise.net.nz) has joined #freenet
[6:03] * nextgens (~nextgens@jabber.hst.ru) has joined #freenet
[6:03] <nextgens> hi
[6:03] <mjr_> yo yo yo
[6:48] * Romster1 (Romster@203.129.145.209) has joined #freenet
[7:11] * Romster (Romster@203.129.151.119) Quit (Read error: 110 (Connection timed out))
[7:14] <sanity> mjr: long time no see
[7:14] <sanity> toad: you awake yet?
[7:14] <mjr_> well, here in lower zimbabwe the internet is still kinda a new thing and so forth
[7:15] <mjr_> weather is nice though
[7:17] <sanity> what have you been up to?
[7:17] <mjr_> worrying about whatever i should be up to
[7:17] <mjr_> my life is still pretty screwed up. hanging out around the house and whatnot
[7:18] <mjr_> but i thought i'd check in and give you all my best new freenet idea while i still want to
[7:18] <mjr_> (which i am presently working on)
[7:20] * mjr_ takes a shower
[7:29] * Ash-Fox (UNKNOWN@ede161.neoplus.adsl.tpnet.pl) Quit (Read error: 104 (Connection reset by peer))
[7:30] * Ash-Fox (UNKNOWN@ede161.neoplus.adsl.tpnet.pl) has joined #FreeNET
[7:35] <mjr_> showers are nice
[7:42] <mjr_> sanity: check out the essay i'm drafting at http://znex.org/alpha.pdf
[7:42] <sanity> will do
[7:44] <sanity> mjr: looks to me like you are taking the simple notion of cover traffic and making it sound very complicated
[7:45] <mjr_> what kind of cover traffic could possibly conceal darknet traffic?
[7:45] <mjr_> i'm just trying to reason my way through a problem that has been bugging me
[7:46] <mjr_> namely, the total ineffectiveness of p2p overlay networks when they are forbidden and easily identified
[7:46] <sanity> well, that depends what the darknet traffic looks like
[7:46] <mjr_> it looks totally different than your typical user's web browsing
[7:47] <mjr_> which is really predictable so you stick out like a sore thumb
[7:47] <sanity> it is true that there isn't much encrypted traffic right now that goes between "leaf" nodes on the Internet
[7:47] <mjr_> and there won't ever be really
[7:47] <mjr_> if there was you could ban it
[7:47] <sanity> but i think, to begin with, you are overestimating the sophistication of the adversary, at least for the moment
[7:47] <mjr_> it would be politically trivial in an authoritarian state
[7:48] <mjr_> because there is like no cost to declaring that individuals can't do anything interesting on the net
[7:48] <sanity> under those circumstances you would need to resort to stego
[7:48] <mjr_> it's terribly hard. the better option is to cleverly replace the internet from under their noses, with a network that isn't so prone to observation
[7:49] <mjr_> and then try to make it politically risky to interfere with (say, by creating an important electronic market that runs on it, or whatever)
[7:50] <sanity> ipv6 perhaps
[7:50] <mjr_> it's just really hard to get much information stego encoded into statistically credible web traffic
[7:51] <mjr_> anything, really, just as long as the network has different alpha-properties (to use my term for the amount of observable information)
[7:51] <mjr_> problem is, such a thing would be opposed, so you have to be sneaky about introducing it.
[7:51] <mjr_> and sneaky about keeping it once it's there (i mean, why not shut it down if all it does is circumvent your censorship)
[7:52] <mjr_> so i am trying to work out a way to force a hostile government to acquiesce. some easy problem....
[7:53] <mjr_> short of the citizen's vote or the assassin's bullet i guess
[7:54] <mjr_> i figure china's communists are acutely sensitive to economic pressures, that is where i got that idea
[7:54] * Hory (~Miranda@82.78.27.85) has joined #FreeNet
[7:54] <mjr_> if the new network is essential for modern business in some way, it would be damn hard to close it down
[8:03] <toad_> mjr_: here in lower zimbabwe?!
[8:03] <mjr_> i think i'll become a hermit fisherman like my uncle tom
[8:04] <mjr_> he is much venerated by my mom
[8:04] <mjr_> you can sing that
[8:05] <mjr_> right after you design the worldwide private computing grid, the new operating system to run on it, the super electronic market system...
[8:05] <toad_> sanity: hi
[8:05] <toad_> mjr_: the worldwide private computing grid in the way you speak of it is a nightmare
[8:06] <mjr_> i know you've got worries about consolidation, but the principle is that you'd have private computers with encrypted private links that make a big private graph
[8:07] <toad_> and that works well if you want to hire some supercomputer power
[8:07] <toad_> but for more general thin-client/storage/etc, for consumers, it doesn't
[8:07] <toad_> because there's a natural monopoly
[8:07] <toad_> if you want to take advantage of the practically zero latency and gigabit+ bandwidth
[8:07] <toad_> you end up with all the world's computers in one place
[8:08] <toad_> owned by one company
[8:08] <toad_> susceptible to one terrorist attack
[8:08] <mjr_> i suppose we can argue the feasibility of grid computing (why not), but it is _totally_ secondary to my argument
[8:08] <toad_> heh
[8:08] <mjr_> i've revised my essay if you care at all. same place
[8:08] <mjr_> my beautiful, beautiful essay with all its logic and greek
[8:09] <mjr_> and so many pages...
[8:17] <mjr_> btw, ever think how hard stego would be when you have to use web pages?
[8:18] * NullAcht15 (~NullAcht1@dsl-082-082-135-193.arcor-ip.net) has joined #freenet
[8:18] <mjr_> the pages are public. you can compare what the page says it is with what it really is
[8:19] <mjr_> i mean, say you had some proxy that encoded messages into public images from web pages
[8:19] <mjr_> the evil tyrant firewall could cache everything and compare what you receive with its cached version
[8:19] * thelema (~thelema@193.220.86.4) has joined #freenet
[8:19] * thelema (~thelema@193.220.86.4) Quit (Client Quit)
[8:20] <toad_> they won't ban SSL
[8:20] <toad_> even in china
[8:20] <toad_> we can hide behind that, but unfortunately that is still very much server-push
[8:21] * Romster1 is now known as Romster
[8:21] * Romster (Romster@203.129.145.209) Quit (Nick collision from services.)
[8:21] * thelema (~thelema@193.220.86.4) has joined #freenet
[8:21] * Romster (Romster@203.129.145.209) has joined #freenet
[8:24] <mjr_> toad_: they may very well ban ssl for individual connections
[8:25] <toad_> mjr_: define individual
[8:25] <mjr_> need an approval letter from the party to get one or something
[8:25] <toad_> yeah they might do that
[8:25] <mjr_> as opposed to business
[8:25] <mjr_> i mean, when you want to control stuff, you control stuff. period.
[8:26] <mjr_> the trick is to make that cost something
[8:26] <mjr_> like a tamper-proof network. you fuck with it, it blows up in your face
[8:26] <mjr_> that's the ideal
[8:27] <thelema> mjr_: just make the party the only issuer of SSL certs.
[8:27] <mjr_> with foresight i suppose you could put a sort of 'wrapper' around anything. i dunno. needs thought
[8:28] <thelema> I plan on putting a BEEP+TLS wrapper around FNP connections.
[8:29] <thelema> toad_: any thoughts re: using git as a datastore?
[8:29] <toad_> thelema: what's git?
[8:30] <toad_> BEEP+TLS? you can do TLS on UDP?
[8:30] <thelema> toad_: linus's content-addressable filesystem.
[8:30] <toad_> lol
[8:30] <toad_> it's a bunch of bash scripts isn't it?
[8:30] <thelema> toad_: no, I'm planning on using TCP.
[8:30] <toad_> will be slow
[8:30] <toad_> but we'll need tcp for most stego
[8:31] <toad_> if the enemy cannot do statistical traffic analysis on individual links, then stego is easy
[8:31] <toad_> just make the conns look like realplayer
[8:31] <toad_> if they CAN, we have serious problems
[8:31] * gvdm (~gvdm@210-246-18-159.paradise.net.nz) Quit (No route to host)
[8:32] <thelema> toad_: Tom Lord (of arch fame/infamy) is doing a bunch in C to make it usable.
[8:32] <toad_> ah cool
[8:32] <toad_> well in 0.7 we have fixed sized files
[8:32] <toad_> which means most content addressible FSs would be overkill
[8:33] <thelema> http://www.seyza.com/=clients/linus/tree/src/liblob/blob.html
[8:33] <thelema> http://www.seyza.com/=clients/linus/tree/src/libdx/blob-xmt.html
[8:33] <mjr_> traffic analysis is kid's stuff
[8:33] <mjr_> my dog does statistical traffic analysis
[8:33] <mjr_> she can also bounce a ball with her snout
[8:34] <toad_> mjr_: no, it's not that easy
[8:34] <toad_> you need MAJOR hardware to do it on real connections
[8:34] <toad_> by real connections i mean the 90Gbps the chinese have
[8:34] <mjr_> it comes out of your isp bill, dude
[8:34] <toad_> or the terabits we have, and which the chinese will have in 10 years
[8:34] <mjr_> the little box at your dsl isp
[8:34] <mjr_> may a thousand boxes bloom, and so on
[8:36] <thelema> mjr_: possible, yes. likely, no.
[8:36] <toad_> brb
[8:36] <mjr_> thelema: when you want control, and you can afford control... you buy the damned control
[8:37] <mjr_> a stego-detecting firewall is not rocket science
[8:37] <thelema> sorry, I have to go.
[8:37] <mjr_> and it is no big deal to have a few engineers slap one together and order all the isps to install them
[8:37] <mjr_> later
[8:37] <thelema> hopefully I'll be back in a few hours
[8:37] * thelema (~thelema@193.220.86.4) Quit ("thelema has no reason")
[8:38] <mjr_> unless a mortar round gets you or something
[8:38] <mjr_> i know it's pretty bad there in iraq these days
[8:41] <mjr_> toad_: bcoates and i did some thinking about big hashtables
[8:41] <mjr_> we broke the hashtable up into 'bins' of N items and did LRU within each bin when a new item came in and an old one needed to go
[8:41] <mjr_> that kept the lru list out of memory
[8:41] <mjr_> while approximating lru replacement
[8:42] <mjr_> that would be how i would stuff a terabyte disk full of kbyte blocks
[8:43] <mjr_> the thing with small blocks is actually disk seeking
[8:43] <mjr_> very slow
[8:44] <mjr_> i never learned the function that defines how long a disk takes to seek, and how it responds to the distance seeked
[8:44] <mjr_> might want to look that one up and see about ordering them at a little latency cost
[8:48] <toad_> mjr_: the point about guns is that if you have good communications you may not need them
[8:48] <toad_> can make the difference between a successful bloodless coup and a rapidly crushed bloody uprising
[8:49] <mjr_> especially if the mortar round hits you first. so how about those seek times
[8:49] <toad_> well.. yet more threads? :<
[8:49] <mjr_> google isn't telling me much
[8:49] <toad_> best way to order them is just to have lots of fetches going at once and let the OS's elevator algorithms order them
[8:50] <toad_> that means either AIO or even more threads
[8:50] <toad_> AIO is only available in java 1.5 :<
[8:50] <toad_> also i'm not sure about kilobyte blocks
[8:50] <mjr_> code it up with the java-c interface
[8:50] <mjr_> oh, i thought you were set on 4kbyte
[8:50] <toad_> there are clear advantages, but there are also clear problems - such as whether we can route at that speed
[8:50] <mjr_> shows how closely i follow freenet
[8:50] <toad_> current theory is 32kB
[8:51] <toad_> also bigger splitfiles make a number of security problems worse
[8:51] <toad_> although those problems are pretty terrible regardless until we get premix routing
[8:51] <toad_> (which is really hard :( )
[8:52] <toad_> but we do need to decide for 0.7 because we really don't want a second network reset
[8:52] <toad_> if we can batch the estimator updates, say updating everything once a minute, then we can route extremely fast
[8:52] <toad_> but during bootstrapping we can't do this, and it may slow things down later on
[8:53] * sanity_ (~ian@81-178-91-152.dsl.pipex.com) has joined #freenet
[8:53] <toad_> we need to simulate that
[8:53] <toad_> if it DOES work it would speed simulations up considerably
[8:53] <toad_> sanity: hi!
[8:54] * linyos (~asdfasdf@adsl-68-79-4-86.dsl.emhril.ameritech.net) has joined #freenet
[8:54] <linyos> some thunderstorm here in equitorial new guinea
[8:54] <toad_> maybe we can have some sort of measure of estimator flux that we use to determine whether to update in batches or not
[8:54] <toad_> linyos: welcome back
[8:54] <toad_> linyos: i thought you were in lower zimbabwe? :)
[8:55] <toad_> sanity_: hi
[8:56] <linyos> can the estimators lag?
[8:56] <toad_> not at the moment
[8:56] <linyos> i mean in theory, would it perversify the routing
[8:56] <toad_> the suggestion is we let them lag so we can precompute routing choices
[8:56] <toad_> well, i think routing may be rather sensitive during the bootstrap phase
[8:56] <toad_> after that, maybe, maybe not
[8:57] * sanity (~ian@81-178-126-192.dsl.pipex.com) Quit (Read error: 60 (Operation timed out))
[8:57] <linyos> the routing table updates can't have that high a complexity factor can they?
[8:57] <linyos> probably just a matter of optimization, writing some C code
[8:57] <toad_> that's not the issue
[8:57] <toad_> and writing C code is not the answer to everything
[8:57] <toad_> the issue is that we may not be able to estimate fast enough
[8:57] <linyos> well, i'm guessing you're doing a whole bunch of expensive floating point
[8:58] <toad_> actually some of it is still BigInteger
[8:58] <toad_> only a small part of it though
[8:58] <linyos> you port the thing to integer arithmetic or simd and it'll speed up like a charm
[8:58] <toad_> we pass keys around as BigInteger's
[8:58] <toad_> well we need to do it in floating point for sufficient accuracy IMHO
[8:58] <toad_> but modern CPUs have really rather nice floating point units
[8:59] <toad_> as far as speed goes...
[8:59] <toad_> lets see
[8:59] <linyos> yeah, i can't see any problem with floats as long as your algorithm has a reasonable complexity
[9:00] <toad_> hmmm i think this simulation may have stopped...
[9:00] <linyos> but you don't need big keys. 64 bits is fine to make brute-force collision attacks infeasible, and you don't care if one of every 2^32 splitfile parts gets is corrupt and has to be erased
[9:00] <toad_> yeah last updated 13:44
[9:00] <toad_> ugh
[9:01] <toad_> no, it's not
[9:01] <toad_> we need big keys
[9:01] <linyos> and why?
[9:01] <toad_> but we cast them to floating point for routing
[9:01] <toad_> because you can make collisions pretty easily with only 64 bits
[9:01] <toad_> because making a collision is only 2^32
[9:01] <toad_> also because the hashing algos are looking decidedly dodgy
[9:02] <linyos> making a collision is pointless because the splitfile reconstructor will verify each part with a larger hash
[9:02] <toad_> meaning we'll probably go to 256 bit keys
[9:02] <toad_> so you can DoS the splitfile
[9:02] <toad_> anyway, this sim stopped at may 7 13:44
[9:02] <linyos> no, DoS isn't a birthday attack. it doesn't have sqrt(bits) complexity, it has bits complexity
[9:02] <toad_> when did it start?
[9:03] <toad_> okay
[9:03] <toad_> even with bits complexity it's looking hazardous
[9:03] <linyos> the birthday attack would mean that you could insert a splitfile where you could swap out all the parts with some other part later on
[9:03] <linyos> which is totally pointless
[9:03] <linyos> the hash algorithm thing is a legit worry
[9:03] <linyos> i'm not happy about that sha1 weakness either
[9:04] <linyos> makes you wish a stronger primitive was available
[9:04] <toad_> argh
[9:04] <toad_> ssh: connect to host cypres3.dyndns.org port 20022: No route to host
[9:05] <toad_> i can't get through to my simulations because nextgens's stupid wireless conn is down!
[9:05] <nextgens> ^-^
[9:06] <nextgens> I'm effraid they've been killed ...
[9:06] <toad_> it looks that way
[9:06] <nextgens> we've had a power failure
[9:06] <toad_> but they may only have been kill -STOPped?
[9:06] <toad_> oh well done
[9:06] <toad_> :|
[9:06] <nextgens> and now, someone is drilling in the room ...
[9:06] <toad_> well do you know how long they were up?
[9:07] <toad_> the ctime isn't available because of the syncing
[9:07] * mjr_ (~asdfasdf@adsl-68-252-188-62.dsl.emhril.ameritech.net) Quit (Read error: 110 (Connection timed out))
[9:07] <toad_> owait
[9:07] <nextgens> since last friday
[9:07] <toad_> 29/04/05 23:48 to 07/05/05 13:44
[9:07] <nextgens> maybe a bit more
[9:07] <toad_> which is a pretty long time
[9:08] <nextgens> as we had a network failure before ^-^
[9:08] <toad_> about 8.5 days
[9:08] * cbreak (~cbreak@84-72-98-199.dclient.hispeed.ch) has joined #freenet
[9:08] <toad_> which is 734,000 seconds or thereabouts
[9:09] <toad_> average path length starts at 4 and goes up to 10 with a log scale
[9:09] <toad_> i don't suppose you have a formula for sigma(log(n))...
[9:12] <toad_> hmmm bash doesn't have floating point
[9:13] * linyos really hopes ian's comment about my overestimating adversaries is right
[9:13] <linyos> cause then i'll definitely take over teh world
[9:13] <toad_> :)
[9:13] <linyos> :) indeed.
[9:14] <linyos> this irc client renders those with yellow icons, it's so cool
[9:14] <toad_> okay, the total number of hops throughout the simulation is...
[9:14] <toad_> 74078.3873514729710536
[9:14] <toad_> times the number of requests
[9:14] <toad_> which is iirc 10,000
[9:14] <toad_> so 740,783,873.5 routing's
[9:14] <toad_> each one having 10 options
[9:15] <toad_> over a time period of 734,000 seconds
[9:15] <linyos> why don't you use genetic algorithms to design the routing function? it could be possible
[9:15] <toad_> on a xeon 2.8GHz
[9:15] <linyos> you evaluate each one with a simulation.
[9:15] <linyos> breed, repeat
[9:15] <toad_> because they usually need even more tweaking than any other algorithm? :)
[9:15] <toad_> woah
[9:15] <toad_> that's one routing per MILLISECOND!
[9:16] <toad_> okay, we can use 1kB blocks, if we sort the security issues out
[9:16] <linyos> not bad, i think the estimator i coded took like 7ms with 2^10 nodes
[9:16] <toad_> well this is with routing between 10 nodes
[9:16] <toad_> this isn't the full estimator formula, but i think the full estimator formula sucks
[9:16] <toad_> and this includes updates too
[9:17] <linyos> excuse me while i laugh at you like a jackass
[9:17] <linyos> :)
[9:17] <toad_> so there is no technical reason CPU-wise why we can't use 1kB blocks
[9:17] <toad_> or even 128 byte blocks
[9:17] <toad_> but that would be crazy because the headers and the key are likely to be close to that
[9:17] <linyos> i think you want to use the whole 14xx byte packet for efficiency's sake
[9:17] <toad_> not for a key
[9:17] <toad_> we need some space for messages too
[9:18] <cbreak> 32k keys are too big then?
[9:18] <toad_> also i'd like to be able to use 660 byte packets to mimic realplayer's traffic
[9:18] <linyos> they get aggregated?
[9:18] <toad_> we COULD use 32kB keys
[9:18] <toad_> or we COULD use 1kB keys
[9:19] <linyos> oh, i thought about this and i decided that freenet needs priority bands for messages
[9:19] <linyos> like, web browsing, you would use immediate-routed messages
[9:19] <cbreak> linyos: that doesn't make sense.
[9:20] <linyos> and downloading a movie, you would use high-latency messages
[9:20] <cbreak> at least not to me :)
[9:20] <linyos> that would be queued for up to 24 hrs
[9:20] <cbreak> since content is all equal.
[9:20] <linyos> cbreak: your demands are certainly not the same.
[9:20] <linyos> sometimes you want low latency, sometimes you don't care
[9:20] <cbreak> my demands are anonymity.
[9:21] <linyos> you get plenty of anonymity either way. it's not really an issue
[9:21] <cbreak> if you don't care about anonymity, why freenet?
[9:21] <linyos> i just mean that the high-latency messages would be queued.
[9:21] <linyos> how does that relate to anonymity?
[9:21] <cbreak> how would you know which messages require which latencity?
[9:22] <linyos> the user's software specifies it based on the application
[9:22] <toad_> freenet does NOT need priority bands for messages
[9:22] <linyos> the movie downloader asks for high latency and the web browser asks for low latency
[9:22] <linyos> toad_: here are the alternatives
[9:22] <toad_> cover traffic requires that any block is virtually indistinguishable from any other block
[9:22] <toad_> otherwise we lose a great deal
[9:22] <linyos> 1) you drop requests pointlessly even though the requester would be happy to wait a while for the reply
[9:23] <linyos> 2) you queue high-priority requests behind low-priority requests for no reason
[9:23] <linyos> toad_: there are two bands. it's one bit. it partitions the requests into two classes.
[9:23] <cbreak> 3) messages are born equal.
[9:23] <toad_> linyos: we intend to make all requests pretty fast as in lowish latency
[9:23] <linyos> each class has plenty of activity at all times.
[9:23] <toad_> with minimal retrying
[9:24] <toad_> and UDP
[9:24] <toad_> IMHO it is easiest to beat traffic analysis with 1kB blocks
[9:24] <toad_> or even 512 byte blocks depending on how big the headers are
[9:24] <toad_> and also it lets the network learn REALLY fast
[9:24] <linyos> what is the objection to the two bands?
[9:24] <toad_> HOWEVER, it exacerbates a number of attacks unless we have premix routing
[9:25] <toad_> linyos: loss of cover traffic. ability to categorize freenet traffic is bad.
[9:25] <linyos> you can categorize it into two stupid categories
[9:25] <linyos> big deal, that is pratically insignificant
[9:25] <cbreak> no.
[9:26] <linyos> and for that price: you get more efficient bandwidth use and less latency all around
[9:26] <toad_> IMHO cover traffic is pretty important
[9:26] <toad_> and especially with 1kB blocks and UDP we can deliver solid latency on every request
[9:26] <linyos> imho you haven't come up with a reason why the single bit of information does anything so harmful
[9:26] <toad_> we have a new load management algorithm which is based on TCP and so should work okay
[9:26] <cbreak> any information is bad.
[9:27] <toad_> cbreak: right
[9:27] <linyos> and one bit of information is negligibly bad
[9:27] <cbreak> no.
[9:27] <linyos> while making the network's throughput and latency much better
[9:27] <toad_> no, one bit of information can be the critical bit you needed as an attacker
[9:27] <toad_> linyos: what exactly is the benefit to setting your requests to low priority?
[9:27] <cbreak> a secure scheme is not allowed to have any corelation between contend and cyphertext.
[9:28] <robilad> hi toad_, sent you a reply :)
[9:28] <linyos> toad_: 1) the public good served by being able to queue them out of the way and let high-prio messages through
[9:28] <linyos> thus dramatically easing network load
[9:28] <robilad> thanks for the mail, btw
[9:28] <cbreak> linyos: it does not ease the load.
[9:28] <toad_> cbreak: right. if your outgoing requests are all low prio, you're probably requesting a splitfile...
[9:28] <cbreak> since freenet uses any available bandwidth anyway :)
[9:28] <toad_> robilad: thanks
[9:28] * Ash-Fox (UNKNOWN@ede161.neoplus.adsl.tpnet.pl) Quit (Nick collision from services.)
[9:29] * Ash-Fox (UNKNOWN@edi80.neoplus.adsl.tpnet.pl) has joined #FreeNET
[9:29] <linyos> cbreak: yeah it does. say the network is totally at capacity with people requesting movies
[9:29] <linyos> cbreak: and i want to view a freesite
[9:29] <linyos> cbreak: guess what happens with your scheme and what happens with mine
[9:29] <robilad> toad_: if it explains things a bit, then I'll send a reply to chat@ too
[9:29] <cbreak> think about all implications before proposing something like that.
[9:29] <cbreak> you wouldn't gain anything.
[9:30] <toad_> robilad: yes please
[9:30] <robilad> i just don;t wnat to do the same work twice :)
[9:30] <cbreak> since priorities would be user specified
[9:30] <cbreak> but you would lose anonymity.
[9:30] <toad_> linyos: load is managed by throttling clients according to a TCP-like algorithm
[9:30] <toad_> actually... oh ****
[9:30] <linyos> cbreak: i'm well aware that users have no direct incentive to set their packets to low priority. but if people are greedy the network explodes anyway.
[9:31] <cbreak> afaik, tcp has a flag for important traffic... which is ignored all throught the internet :)
[9:31] <linyos> toad_: and, no, you decrease anonymity by a very small constant factor
[9:31] <linyos> i mean cbreak
[9:31] <toad_> linyos: people are. you have to build feedback in so that it doesn't matter.
[9:31] <cbreak> if the question is anonymity or something else, the answer is Anonymity.
[9:32] <toad_> cbreak: within reason
[9:32] <linyos> toad_: then freenet blows up, because when many users turn greedy all sorts of horrible things happen
[9:32] <toad_> but certainly in this case
[9:32] <cbreak> (unless something else is lot's of money )
[9:32] <linyos> they really start fucking up the network
[9:32] <toad_> lol
[9:32] <cbreak> not more than now :)
[9:32] <toad_> linyos: load balancing is a problem
[9:32] <toad_> it has always been a problem
[9:32] <linyos> toad_: priorities solve that problem.
[9:32] <linyos> two: high and low.
[9:32] <toad_> ian has proposed a tcp-derived way to deal with it for 0.7
[9:32] <toad_> linyos: i don't think so
[9:33] <toad_> they give away way too much information
[9:33] <linyos> i figure the network capacity demanded by high-priority requests for freesite pages is very low compared to what's demanded by huge downloads
[9:33] <toad_> if there's no other way to solve the load problem then we'll consider it
[9:33] <linyos> toad_: they give away one bit of information
[9:33] <toad_> yes and that one bit can be vitally important sometimes
[9:33] <linyos> i can't see any big consequence to that.
[9:34] <toad_> especially if you have thousands of them
[9:34] <linyos> so this is an application of the "precautionary principle" -- refusal to consider new things because you might be wrong about their effects
[9:34] <toad_> my concern today, and thanks for pointing me in the right direction, is that ian's tcp-inspired load balancing algorithm will give away the client node...
[9:35] <toad_> linyos: we may have to rethink the load balancing thing yet again
[9:35] <toad_> linyos: in which case we will have to reconsider your idea
[9:35] <cbreak> linyos: it's reasonable: don't introduce new and potentialy dangerous things unless they are absolutely neccessary.
[9:35] <linyos> cbreak: i'd say fixing freenet's horrid latency is pretty necessary, wouldn't you?
[9:35] <toad_> i hope that we can avoid actual delaying queueing, and stick with some form of unobtanium-based simulated queueing
[9:35] <linyos> it's always been nearly unusable.
[9:36] <cbreak> freenet's latency isn't a problem for me :)
[9:36] <linyos> queueing is totally rational. it's a natural way to smooth out an irregular demand
[9:36] <cbreak> and too, 0.7 will be completely new.
[9:36] <cbreak> you can't say anything about it's performance.
[9:36] <toad_> but if we do have to queue for 1000ms say on each node, with 1kB blocks and a max htl of around 30 including the random bits, that wouldn't be SO terrible
[9:37] <linyos> certainly dropping requests and therefore making people hammer the network over and over is no better
[9:37] <toad_> linyos: i mean the queueing we do in 0.6
[9:37] <toad_> linyos: not priority based queueing, different queueing
[9:37] <toad_> linyos: well... there's a trade-off. if you keep on retrying forever you multiply the input load
[9:37] <toad_> if you don't, you force the client to retry
[9:37] <toad_> the latter is on the whole preferable
[9:38] <linyos> i'm no fan of retrying.
[9:38] <toad_> if average successful htl is 20 (on a big network), and we queue for 500ms, and we have a 200ms transit time... then we get 700*20 = 14 seconds typical latency
[9:38] <toad_> hmmmm
[9:38] <linyos> i say let stuff drop and have the clients sort out the mess
[9:39] <linyos> makes everything loads simpler
[9:39] <toad_> yes
[9:39] <toad_> but we probably don't need to actually queue
[9:39] <toad_> we do simulated queueing
[9:39] <toad_> on input, we do the estimates for the request against our nodes
[9:40] <linyos> there is a fundamental issue with using the same connections for messages with differing latency demands
[9:40] <linyos> you simply cannot do anything intelligent unless you have a bit that tells you how to prioritize
[9:40] <toad_> if it's in the bottom X% of requests by estimate rank (X depends on load, ideally 0), we kill it then
[9:41] <linyos> and with the network chock full of movie downloads, there is going to be tons of failure and delay
[9:41] <toad_> then we send it out immediately
[9:41] <toad_> problem is that we can't have it sent out immediately if we have MRIs
[9:41] <linyos> popularity has nothing to do with latency requirements
[9:41] <toad_> and if we don't have MRIs we have to have something else...
[9:41] <linyos> a popular block could be a part of a big movie, an unpopular block could be part of a web site
[9:42] <toad_> client-end throttling as ian suggests is probably a really bad idea because it probably gives away who the client is
[9:42] <toad_> which isn't a problem for TCP
[9:42] <toad_> but is a HUGE problem for freenet
[9:42] <toad_> building a scalable anonymous filesharing network is so hard... :|
[9:42] <toad_> still i'm glad i caught that one before we implemented it
[9:42] <linyos> you don't have to convince me about that. i decided it was impossible
[9:43] <toad_> thanks for putting me onto it
[9:43] <linyos> you're very welcome.
[9:43] <toad_> well, we may consider prioritizing blocks, but it's a last resort IMHO
[9:44] <linyos> you should discuss how much it decreases anonymity
[9:44] <toad_> i don't see that it will solve the load balancing problem
[9:44] <linyos> it won't
[9:44] <toad_> and it looks to me like the incentives are all set up so that everyone will set the flag to high priority
[9:45] <toad_> on every request
[9:45] <linyos> it solves the problem of keeping important requests working at the expense of unimportant requests
[9:45] <cbreak> problem: important is user controlled.
[9:45] <linyos> toad_: yeah, but like i said, if users are all hot about gaming the system, it will collapse whether or not you have priorities.
[9:45] <toad_> if we have 1kB blocks, and a 10kB/sec uplink... then we will on average have one request every 100ms. Divided by the psuccess.
[9:45] <toad_> so at the most, one every 100ms.
[9:45] <linyos> and btw, if everyone sets high priority it's just like you have no priorities
[9:46] <toad_> possibly one every 10ms, but only if the network sucks
[9:46] <linyos> cbreak: if only 5% of freenet users are willing to be antisocial, they will probably flood the network to hell
[9:46] <toad_> linyos: you have to set the incentives the right way
[9:46] <toad_> and i don't see how you can do that and maintain anonymity
[9:46] <cbreak> linyos: there will be more than 5%. The RIAA, CIA, MI5, ...
[9:46] <linyos> toad_: yeah, and stuffing that into freenet is a bitch
[9:47] <toad_> well maybe with premix routing
[9:47] <toad_> if you have one request every 50ms, say
[9:47] <linyos> cbreak: that's why freenet is doomed because of its inefficiency and zero cost
[9:48] <toad_> taking into account frost etc
[9:48] <cbreak> zero cost? tell that my provider :)
[9:48] <linyos> flooding has to be the nastiest intractable problem with freenet
[9:48] <toad_> then if you queue for 300ms, you have six requests to choose between in that period
[9:48] <linyos> cbreak: i mean you can get other nodes to wear themselves out without doing much work or compensating them in any way
[9:48] <toad_> linyos: intentional flooding is not the problem right now
[9:48] <toad_> it's load balancing in general
[9:48] <linyos> toad_: then tomorrow it'll be the RIAA flooding you to hell
[9:49] <linyos> it is just inevitable
[9:49] <toad_> hmmmmmmm
[9:49] <toad_> this is rather disturbing
[9:50] <toad_> if you have one request every 50ms, then a plausible MRI might be...
[9:51] <toad_> lets see, right now we have say a 10kB/sec uplink with an average file size of 400kB if successful
[9:51] <toad_> and a success ratio of maybe 5%
[9:51] <toad_> that makes 400kB/10kB/sec = 40 secs, times 5% = 2 secs
[9:51] <linyos> MRI? mean request interval?
[9:51] <toad_> actual MRIs vary wildly...
[9:51] <toad_> minimum request interval
[9:52] <linyos> ok, a bandwidth limit
[9:52] <toad_> not exactly
[9:52] * FallingBuzzard (~srademach@66.151.22.70) has joined #freenet
[9:52] <toad_> a means of load balancing we developed over the last couple years
[9:52] <linyos> oh, you mean you developed something without my invaluable guidance?
[9:52] <linyos> how did that happen.
[9:52] * toad_ needs to connect to his node to get some real MRIs...
[9:52] <toad_> LOL
[9:53] <toad_> you have a node?
[9:53] <toad_> what's a typical MRI?
[9:53] <linyos> nah
[9:53] <toad_> of course in a fixed mesh the number of connections would be way lower...
[9:53] <linyos> i'm too good to run a node, what with all the pedos and so forth running amok
[9:54] <toad_> heh
[9:54] <linyos> not to mention osama
[9:54] <linyos> load balancing is tricky
[9:54] <toad_> i suppose MRI is a function of the average interval between requests and the level of mal-distribution of load on the network
[9:55] <linyos> because if it's not a part of the routing algorithm, you start skewing the routing.
[9:55] <toad_> if it IS a part of the routing algorithm you skew the routing
[9:55] <linyos> i mean then it would be taken into account, you'd know the routing works
[9:55] <toad_> you end up routing requests to the least overloaded node
[9:55] <toad_> MRI + queueing was a sort of compromize that showed some promise
[9:55] <linyos> as opposed to filtering your routing decisions through some external filter that does god knows what to their optimality
[9:56] * NullAcht15 (~NullAcht1@dsl-082-082-135-193.arcor-ip.net) Quit (Remote closed the connection)
[9:56] <toad_> my most popular node has an MRI of 7 seconds
[9:56] <toad_> the next one has an MRI of 190 seconds
[9:56] <toad_> the next one has 266
[9:56] <toad_> then it's 9
[9:56] <toad_> 300
[9:56] <toad_> 105
[9:57] <toad_> 192
[9:57] <toad_> 1
[9:57] <toad_> 18
[9:57] <toad_> 300
[9:57] <toad_> they're all over the place, basically
[9:57] <toad_> and mostly well over the maximum queueing time
[9:58] <toad_> which depends on the file size
[9:58] <toad_> curiously enough
[9:58] <toad_> sort of mimicing your priorities :)
[9:58] * thelema (~thelema@193.220.86.4) has joined #freenet
[9:58] <linyos> i thought you moved to constant filesize
[9:58] <toad_> local requests are queued for 1 minute
[9:58] <toad_> linyos: working on it
[9:59] <toad_> 0.7 will
[9:59] <toad_> remote requests are 10 secs/megabyte, with a minimum of 3 seconds
[9:59] <toad_> (double for inserts)
[9:59] <toad_> now
[9:59] <toad_> if we cut the typical time between requests from 2 seconds to 50 milliseconds
[10:00] <linyos> oh well, i was out of touch even before i threw away my computer
[10:00] <toad_> and cut the typical number of nodes from say 50 to say 5
[10:00] <toad_> that ought to reduce the average MRI by a large factor, say 400 or so?
[10:00] <thelema> yes, it should...
[10:01] <toad_> in which case a queue time of 1000ms should be adequate for most nodes
[10:01] <toad_> unfortunately a real life fixed links topology would probably have really bad load distribution
[10:01] <thelema> if you cut the time between requests *completing*
[10:01] <toad_> maybe worse than we have now
[10:01] <toad_> thelema: well it doesn't matter, it's the rate of requests being started
[10:01] <linyos> have you gotten closer to deriving a routing algorithm from that 1/d guy's routing proof?
[10:02] <cbreak> toad_: if a node with 50 connections has a mri of 1 sec, does that mean he gets at most 1 request per second per node?
[10:02] <toad_> linyos: oskar thinks that LRU follows 1/d
[10:02] <cbreak> or with mri 50 ms -> 1000 requests per second.
[10:02] <toad_> cbreak: you have MRIs for each node
[10:02] * greycat (rfc1413@wooledge.org) has joined #freenet
[10:02] <toad_> cbreak: but essentially
[10:02] <linyos> standard freenet routing with LRU
[10:02] <thelema> I'm mostly worried about the rate that requests complete.
[10:03] <toad_> thelema: that's the same rate as the rate they are started at
[10:03] <toad_> thelema: logically
[10:03] <toad_> thelema: assuming the system is stable :)
[10:03] <thelema> that's exactly the point I'm making.
[10:03] <thelema> freenet hasn't been stable, because the completion rate has been slower than the starting rate.
[10:04] <toad_> thelema: depends what you mean by stable. the starting rate has to be equal to the completion rate. in order to maintain this we either reject requests or increase our MRI.
[10:04] <toad_> either way is pretty bad for routing
[10:04] <dopeic> this is mabe litel of topic, anybode runing bitchx and can say wath syntax it was to create a new window ?
[10:04] <linyos> so load balancing might be to use that algorithm but have each node massage its datasource reset frequency
[10:04] <greycat> ask #bitchx
[10:04] <linyos> probably /window new
[10:05] <dopeic> the channel dosent really excist :(
[10:05] <toad_> linyos: we already do that
[10:05] <toad_> it's not enough
[10:05] <greycat> maybe everyone but you and the one person there has *finally* decided to scrap bitchx for irssi.
[10:05] * dopeic (ice@c-38c570d5.016-48-6f72652.cust.bredbandsbolaget.se) Quit ("BitchX: No windows left!")
[10:05] <linyos> toad_: it should be enough. everything says the request load coming into a node is a function of its datasource propagation
[10:05] <linyos> if that's not true, something is horribly wrong
[10:06] <toad_> thelema: given that most of the MRIs are huge right now, much smaller keys ought to improve the performance of queueing radically
[10:06] <toad_> linyos: no, it isn't
[10:06] <linyos> i guess the network might be too dynamic in its routing preferences
[10:06] <toad_> linyos: no, the network is too static
[10:06] <toad_> if we ALWAYS used the datasource, then maybe it would be
[10:06] <linyos> like, one day one set of datasources gets hammered for some reason
[10:06] <toad_> if we stick to our nodes and try to learn about them, then we will route a lot of load to a node if it is good
[10:06] <thelema> toad_: the network is too static?!
[10:06] <toad_> even if it is overloaded
[10:07] <toad_> thelema: the network is too static for oskar's results on LRU to really apply
[10:07] <linyos> hmm
[10:07] <toad_> thelema: or for linyos's intuition about datasource resets
[10:07] <toad_> well given NGR
[10:07] <toad_> if we had classical routing maybe it'd work
[10:07] <toad_> but i'd be really disappointed if we had to go back to classical routing
[10:08] <linyos> yeah, right, you route to the datasource 1/30th of time, otherwise to the nearest node to the key in your table.
[10:08] <linyos> (in classical routing, more or less)
[10:08] <linyos> or is that wrong, it's been a long time
[10:08] <toad_> linyos: not sure what you mean
[10:08] * NullAcht15 (~NullAcht1@dsl-082-082-135-193.arcor-ip.net) has joined #freenet
[10:09] <linyos> ok that's probably wrong. ignore me for a moment while i come to terms with having forgotten the freenet algorithm
[10:09] <toad_> hmmm
[10:09] <toad_> i remember thinking that actually cutting the keysize won't reduce the MRI that much
[10:09] <toad_> is that wrong?
[10:09] <toad_> logically if you cut the keysize, you reduce the time between started requests
[10:09] <thelema> linyos: that's the old algorithm (= classical routing/CPalgo)
[10:09] <toad_> you get more requests of course
[10:09] <toad_> but it should reduce the MRI, yes? thelema?
[10:10] <thelema> nope. sorry.
[10:10] <toad_> thelema: huh?
[10:10] <thelema> if keys are half as big, and there's twice as many requests, the MRI is the same.
[10:10] <toad_> hmmm. why?
[10:10] <toad_> MRI is the interval between requests
[10:11] <toad_> so if the keys are half as big, and a node's MRI is the same, then it's going to do half as much work
[10:11] <toad_> so logically the MRI must be halved for it to do the same amount of work
[10:11] <toad_> if we consider the connection between A and B in isolation
[10:11] <thelema> oh, excuse me. yes, the MRI will be halved.
[10:11] <thelema> my bad.
[10:11] <toad_> which we can do because the grid is the same
[10:11] <toad_> okay
[10:11] <toad_> so psuccess increases from 5% to 50%
[10:11] <toad_> and key size reduces from 400kB to 1kB
[10:11] * NullAcht15 (~NullAcht1@dsl-082-082-135-193.arcor-ip.net) Quit (Remote closed the connection)
[10:11] * dopeic (ice@c-38c570d5.016-48-6f72652.cust.bredbandsbolaget.se) has joined #freenet
[10:12] <thelema> 1K keys?
[10:12] <toad_> result: MRI reduced by a factor of 400/10 = 40
[10:12] <toad_> yup
[10:12] <toad_> it's quite possible, we can route in a millisecond
[10:12] <linyos> btw, is there really any basis whatsoever for saying the classical algorithm is a failure? the network has always been so dysfunctional that it would be hard to read anything out of its behavior.
[10:12] <thelema> in simulation, right?
[10:12] <toad_> a bit more with queueing and an extra hash
[10:12] <toad_> right
[10:12] <toad_> linyos: NGR solves a variety of problems/attacks which we can't solve with classical routing...
[10:13] <linyos> ok, i can buy that
[10:13] <toad_> linyos: that doesn't mean it works though!
[10:13] <linyos> (which still leaves the zillion insoluble problems)
[10:13] <toad_> i need to run some more NGR+LRU simulations
[10:13] <toad_> i think i am wrong about NGR+LRU because i haven't been simulating growing networks
[10:13] <thelema> In a content-addressable keyspace, I'd rather not have that many keys. There's a lot of people attacking git for this.
[10:14] <toad_> thelema: hmm?
[10:14] <thelema> I"m not too worried about LRU. I approve of a slowly-changing connection topology.
[10:14] <thelema> back to git,...
[10:14] <toad_> thelema: there are correlation attacks which are exacerbated by having small keys
[10:14] <toad_> but you _can_ do them anyway
[10:15] <toad_> unless we have premix routing
[10:15] <toad_> which is HARD
[10:15] <thelema> we're not going to do 1K keys, right?
[10:15] <linyos> what's wrong with lots of keys?
[10:15] <toad_> thelema: currently the party line is 32kB keys
[10:15] <toad_> thelema: I'm playing with the idea of 1K keys
[10:15] <linyos> and the correlation attacks will work just fine even if you don't have splitfiles
[10:16] <thelema> Well, I'm against 1K keys. Not for the reason of correlation, but for the reason of efficiency
[10:16] <linyos> because whole files can be correlated. say, by being images in a single web page.
[10:16] <thelema> both of individual requests and of splitfile data.
[10:16] <toad_> several advantages: 1. routing learns REALLY fast. 2. we don't have to CBR pad to defeat traffic analysis (we do if keys are > packet size).
[10:16] <toad_> probably there are a whole bunch of others
[10:16] <toad_> oh yeah
[10:16] <linyos> yeah, the splitfile reconstruction complexity can get pretty nasty
[10:16] <toad_> 3. We can use MRIs and queueing, with short queue times and miniscule MRIs
[10:16] <linyos> try keeping a bipartite graph for a billion keys in memory
[10:17] <toad_> thelema: correlation is much harder with fewer blocks to correlate... and it's much harder with linked files rather than splitfiles
[10:17] <toad_> linyos: so segment them
[10:17] <thelema> linyos: feh. that's not so hard.
[10:17] <toad_> thelema: I think ian's idea for TCP-based load balancing will give away the client
[10:17] <thelema> toad_: you're suggesting having more blocks.
[10:18] <toad_> thelema: so we'll have to stick with MRIs
[10:18] <toad_> MRIs work best if they are low enough for queueing to be actually useful
[10:18] <toad_> thelema: indeed
[10:18] <toad_> thelema: we had this debate a while ago, ian was opposed, ian won
[10:18] * linyos is confident that thelema really knows good algorithms for assembling billions of blocks
[10:18] <toad_> but i think it may need to be gone through again
[10:18] <toad_> err
[10:18] <toad_> ian opposed small keys
[10:19] <linyos> xor-folding a hash into 64 bits doesn't make algorithmic attacks any easier does it?
[10:19] <thelema> CHK is a content-expansion trick. He realized this back in the beginning. The freenet network becomes a decompression device for CHKs into files.
[10:19] <thelema> be back in a few minutes.
[10:19] <thelema> (sorry)
[10:20] * thelema (~thelema@193.220.86.4) Quit ("thelema has no reason")
[10:20] <toad_> hmmm
[10:20] <toad_> okay
[10:20] <toad_> be back in half an hour, okay?
[10:20] <linyos> sure thign
[10:20] <toad_> is there an urgent reason for me to stay?
[10:20] <toad_> bbiab
[10:20] <linyos> watch out for mortar rounds
[10:34] <linyos> if there is one thing you freenet devs have, it's perseverance. i mean, there are like seven intractable problems hanging over your heads yet you keep hammering away and hoping for a deus ex machina
[10:34] <linyos> it's amazing
[10:34] <linyos> it's scary
[10:34] <toad_> :)
[10:35] <greycat> toad_ is better than a million monkeys
[10:35] <toad_> greycat: thanks :)
[10:38] * linyos returns to the ever so fun task of trying to figure out what kinds of network technology can be politically risky to interfere with
[10:40] * sanity_ (~ian@81-178-91-152.dsl.pipex.com) Quit (Remote closed the connection)
[10:44] * sanity (~ian@81-178-91-152.dsl.pipex.com) has joined #freenet
[10:44] * sanity (~ian@81-178-91-152.dsl.pipex.com) Quit (Remote closed the connection)
[10:48] * sanity (~ian@81-178-91-152.dsl.pipex.com) has joined #freenet
[10:48] * sanity (~ian@81-178-91-152.dsl.pipex.com) Quit (Remote closed the connection)
[10:51] * nextime (~nextime@213-140-6-96.fastres.net) Quit (Read error: 104 (Connection reset by peer))
[10:52] * nextime (~nextime@213-140-6-96.fastres.net) has joined #freenet
[10:53] * sanity (~ian@81-178-91-152.dsl.pipex.com) has joined #freenet
[10:54] * sanity (~ian@81-178-91-152.dsl.pipex.com) Quit (Remote closed the connection)
[10:55] * sanity (~ian@81-178-91-152.dsl.pipex.com) has joined #freenet
[11:00] <linyos> what worries me: [TCPA-net] <-> [government proxy] <-> [network at large]
[11:00] <linyos> one day you just say, time to trade in your old network connection for this nice new one...
[11:01] <linyos> gotta make that hard
[11:06] <linyos> what kind of relationship could a user have with the network where you'd say, "oh, we can't do that through a proxy"
[11:08] <linyos> seems impossible
[11:18] <linyos> if it comes to that, you just watch everybody's computer screens with TCPA
[11:18] <linyos> doesn't matter how encrypted your connection is then
[11:19] <linyos> this god damned awful world
[11:19] <toad_> you don't have the manpower, and you'll never entirely stamp out black market hardware
[11:19] <toad_> or the interface from that to the real network
[11:20] <linyos> nah, it could be done in an authoritarian country.
[11:20] <linyos> even in a (formerly) free one.
[11:20] <linyos> you just make ordinary internet connections illegal
[11:20] <linyos> and replace them with TCPA connections
[11:21] <linyos> which have the built-in screen readers
[11:21] <linyos> you know, the new internet national security program, blah blah
[11:23] <linyos> TCPA hardware is hard as hell to break
[11:23] <linyos> like a jail cell for computing
[11:24] <linyos> and as i love to say, where there's a political will, there's a way
[11:25] <linyos> problem is how to break that will
[11:25] <gregh> the rejection of the fcc broadcast flag bodes well for individuals who want to retain rights over their computing
[11:25] <linyos> and my latest idea just failed
[11:25] <linyos> gregh: yeah, for now, and only in the US.
[11:25] <gregh> true. but it's a start. :)
[11:27] <linyos> mind you, i think "everybody move to the US" is a great idea
[11:27] <linyos> or germany, for that matter
[11:27] <linyos> or taiwan
[11:28] <toad_> heh
[11:28] <linyos> but as that's not likely to happen any time soon, we have got to contend with these authoritarian jackasses in every which place
[11:29] <linyos> you could fit earth's entire population into a 50 mile cube with plenty of living space
[11:30] <toad_> including the US :)
[11:31] <linyos> all relative. north korea is a pretty decent place if you ignore just a few things and tweak a few of your values.
[11:35] <toad_> uh, things like extreme poverty?
[11:35] <toad_> completely ignoring the totalitarian rule
[11:36] <linyos> yeah, ignore the famine and develop an admiration for some aspect of the dear leader, combine that with an animosity towards liberalism or something
[11:36] <linyos> you'd get along just fine
[11:37] <linyos> you would find ways to satisfy your philosophical needs with what's there
[11:37] <toad_> and you might have to re-orientate your philosophical and religious views too - certainly in china there are problems with registration of churches
[11:37] <toad_> linyos: the powers of rationalization of the human mind are quite scary really
[11:38] <linyos> do your views serve your emotional needs or vice versa? good question
[11:38] <toad_> my emotions and my reason serve my values
[11:38] <toad_> or they should
[11:38] <linyos> you'd want to find something in kim jong-il and eventually you'd find something.
[11:38] <linyos> yeah, i'm just being cynicall
[11:41] <toad_> "all that is required for evil to triumph is that good men do nothing..."
[11:41] <linyos> but it triumphs anyway
[11:41] <linyos> that is the constant
[11:41] <toad_> sometimes evil triumphs
[11:41] <toad_> not always
[11:41] <linyos> said ivan karamazov
[11:42] <toad_> sometimes good wins
[11:42] <toad_> e.g. the abolition of slavery
[11:42] <linyos> but the victory is perverse and meaningless in a world of such senseless suffering.
[11:43] <linyos> that was ivan's thesis
[11:44] <toad_> so he was a nihilist then
[11:44] <toad_> ?
[11:45] <toad_> surely if you have a totally nihilist worldview then the only logical thing to do is alter your values to be totally cynical and self-centered?
[11:45] <linyos> i'll dig up the paragraph
[11:45] * thelema (~thelema@193.220.86.4) has joined #freenet
[11:46] <thelema> sorry for the longer-than-expected away
[11:46] <toad_> unless you aren't actually all that nihilist - you might think the above but have the caveat that it's worth fighting anyway because you'll be rewarded in the afterlife, for example
[11:46] <toad_> thelema: hi
[11:46] <linyos> It???s not God that I don???t accept, Alyosha, only I most respectfully return him the ticket.??? ???That???s rebellion,??? murmered Alyosha, looking down. ???Rebellion? I am sorry you call it that,??? said Ivan earnestly. ???One can hardly live in rebellion, and I want to live. Tell me yourself, I challenge your answer. Imagine that you are creating a fabric of human destiny with the object of making men happy in the end, giving
[11:46] <linyos> there ya go
[11:46] <greycat> thelema: as you can see, the discussion has been entirely technical and freenet-related.
[11:46] <linyos> if all that got through
[11:47] <greycat> linyos: chopped at "in the end, giving"
[11:47] <toad_> heh
[11:47] <linyos> them peace and rest at last, but that it was essential and inevitable to torture to death only one tiny creature ??? that baby beating its breast with its fist, for instance ??? and to found that edifice on its unavenged tears, would you consent to be the architect on those conditions? Tell me, and tell the truth.??? ???No, I wouldn???t consent,??? said Alyosha softly.
[11:47] <linyos> that is the best statement of the 'problem of evil' i have ever seen
[11:48] <thelema> greycat: yes, of course.
[11:49] <linyos> jesus, i get depressed about networking and i go right to dostoevsky
[11:49] <toad_> linyos: it's been done. a zillion and one times. all the political utopias mess up. the problem is not really structural, it's that people are lazy, greedy, selfish and generally trouble
[11:50] <linyos> it's not about politcs. it's challenging god's decision to create the world..
[11:50] <toad_> linyos: you can't fix the world without fixing the people. and jesus is exactly how you do it. but lets get back to vaguely on topic, shall we?
[11:50] <linyos> you mean this isn't #dostoevsky? my bad...
[11:51] <linyos> back to worrying about networks and political intervention and so on
[11:51] <toad_> :)
[11:51] <toad_> not surprisingly, ian is not exactly keen on 1K blocks
[11:54] * Romster1 (Romster@203.129.145.209) has joined #freenet
[11:55] <toad_> hmmm
[11:55] <toad_> soz ppl gotta bbiab another 20 minutes or so
[11:56] <linyos> so how could you possibly impose a political cost on the decision to install TCPA-monitors everywhere... and if you can't, why bother trying when all you can hope for is a transient victory?
[11:56] <linyos> god this bugs me
[11:56] <linyos> if only i could hang on to a nice answer to this problem
[11:57] <linyos> i would feel so much better
[11:57] * Romster (Romster@203.129.145.209) Quit (Nick collision from services.)
[11:57] * Romster1 is now known as Romster
[11:57] <greycat> linyos: the only way I can see would be to bring people to a state where they're no longer led, sheep-like, to support anything in the name of being safe from Terror(tm).
[11:58] <linyos> that's what i hate about this line of reasoning. it heads straight to the darth vader solution: the only way is to acquire political power.
[12:00] <linyos> i wish the freenet premise was more successful
[12:01] <linyos> i mean, sure, there is plenty of scope for privacy improvements in, say, the US political milieu. but that is a boring challenge
[12:02] <linyos> i've always wanted to stand up for freedom where it is actually challenged, in this project
[12:02] <toad_> linyos: here's some baselines: 1. freenet will never be perfect. it will never be impossible for a sufficiently motivated government to shut it down. 2. it will not be the last word in anonymous p2p and may be part of a larger system, or a step before it. 3. we want something that is better than what is available now, we don't need instant perfection.
[12:03] <linyos> it just doesn't mean anything when the latest ruler decides he's had enough of free networks and pushes the button
[12:03] <toad_> linyos: there's loads of ways you can help to prevent the widespread adoption of TCPA
[12:03] <thelema> toad_: wow. from your mailing lists, I thought you would quit on freenet is any of 1-3 were true.
[12:04] <thelema> s/is/if/
[12:04] <toad_> thelema: it may still be possible to stamp out a scalable darknet
[12:04] <toad_> it will just be expensive
[12:04] <toad_> hopefully you can keep on improving it, and sooner or later the cost becomes a serious concern
[12:05] <linyos> it's easy to lock down the internet completely. you install government servers and do everything through them..
[12:05] <linyos> could be done tomorrow
[12:05] <toad_> linyos: for several trillion dollars, and a huge political loss, and a major economic loss, sure
[12:05] <linyos> browse the web with IP routing, browse the web through the govt proxy, what's the difference
[12:05] <linyos> (except all the censorship and surveillance and so forth)
[12:05] <toad_> depends what you mean by everything i suppose
[12:06] <toad_> but they can surveil you whether you use a proxy or not
[12:06] <toad_> loads of ISPs use transparent proxies
[12:06] <linyos> i'm talking about your typical internet user who wants to read the news or write an email
[12:06] <toad_> which is scary, because it means such a thing is feasible, which means that just about anything is feasible in terms of packet mashing...
[12:06] <toad_> a LARGE fraction of internet users use filesharing
[12:06] <linyos> putting him behind some kind of proxy costs very little and disrupts basically nothing
[12:06] <linyos> except his freedom
[12:07] <toad_> which is technically illegal, because they use it to download illegal files
[12:07] <linyos> i would just like to do more than hope that some jackass doesn't push the button tomorrow or the next day or next year
[12:07] <linyos> to change that jackass's calculus
[12:08] <linyos> that would be a real achievement
[12:08] <linyos> and i almost thought i had a way
[12:08] <toad_> well, what are you saying? everyone gets NATted, you can only run a server with government approval?
[12:08] <linyos> sure
[12:08] <toad_> it might be possible in china but the cost would be enormous
[12:08] <toad_> they're already struggling to keep up
[12:08] <toad_> i think they will keep up enough to do serious damage
[12:08] <linyos> the cost of installing a NAT at an isp. that's real big, sure, sure
[12:09] <toad_> especially with the vague rules they have, and the possibility of extreme sanctions
[12:09] <toad_> linyos: no
[12:09] <toad_> linyos: the cost of licensing every single legitimate server anywhere in china
[12:09] <toad_> right now they only have to license web chat forums and so on
[12:09] <linyos> just let business internet traffic through unfiltered then
[12:09] <toad_> sites where people can post
[12:09] <toad_> linyos: it's not that easy to distinguish between the two
[12:09] <linyos> it's not like that will do me any good at the internet cafe or in my bedroom through my residential connection which is filtered
[12:09] <toad_> linyos: human beings run businesses
[12:10] <linyos> so you get 95% of the problem
[12:10] <linyos> a few servers in some closet are free
[12:10] <toad_> thelema: I don't see much point working on freenet if the chinese can shut it down for $500/year
[12:10] <linyos> an IT guy gets to read the news
[12:10] <linyos> real scary
[12:10] <toad_> thelema: if it costs them $50,000,000/year to shut it down, then there is some point to it
[12:11] <toad_> if it costs them $1,000,000,000/year, then that's a great reason to work on it!
[12:11] <thelema> toad_: reasons 2 and 3 you just gave aren't good enough?
[12:11] <toad_> thelema: reason 2 is a reason TO develop freenet
[12:11] <toad_> it's part of a process
[12:11] <toad_> there will be better systems
[12:11] <toad_> and they will learn from our failures and our successes
[12:12] <linyos> they will get nowhere of consequence unless they find a way to deter rulers from pushing the lockdown button
[12:12] <linyos> to make that decision politically costly
[12:12] <toad_> one way to make it politically costly is to make it financially costly to the government
[12:12] <linyos> to oppose that decision to the rulers' interests
[12:12] <toad_> another way to make it politically costly is to make it a significant cost to the economy
[12:13] <linyos> and i think you can lock things down pretty tight without costing much at all
[12:13] <toad_> a third way is to make it so that it would really piss the people off
[12:13] <linyos> now we're getting somewhere
[12:14] <linyos> my idea for today if you don't remember was along these lines. the network would serve all sorts of vital economic purposes and interfering with it would grind the economy to a half, enraging the rich and poor alike
[12:15] <linyos> what broke that one was that it'd be pretty easy to, instead of interfering in the network, just wrap the access to it with some proxy
[12:15] <toad_> linyos: well, i can see some significant non-controversial uses for freenet, such as free advert-free webhosting of big popular files
[12:15] <toad_> linyos: but only if it's fast
[12:15] <thelema> toad_: fast and efficient
[12:15] <toad_> only relevant in the US though
[12:16] <linyos> those uses don't really interest me. there are fourteen interesting computing problems i could work on if i ever had the urge
[12:16] <linyos> user interface principles, grid computing, to name a couple
[12:16] <toad_> linyos: i'm suggesting that if it was popular enough the US government might not want to attack it
[12:16] <linyos> ok
[12:17] <linyos> along these lines
[12:17] <toad_> linyos: grid computing for the masses is a white elephant. if your vision ever got implemented it would mean we all rent computers hosted by microsoft at one datacenter somewhere in redmond. it would be a nightmare.
[12:17] <linyos> how do you get people to love a computer network?
[12:17] <toad_> and it would likely get bombed at some point
[12:17] <toad_> linyos: how much of the online population of the US downloads illegal files from kazaa, bittorrent etc?
[12:17] * linyos will not argue the grid thing now
[12:18] <linyos> would they take it to the streets if p2p was suppressed?
[12:18] <linyos> would they make it a voting issue?
[12:18] <linyos> i don't know
[12:19] <toad_> probably not because they think it's wrong and do it anyway, on the whole
[12:19] <toad_> so they won't stick up for it politically because they know they don't have a leg to stand on
[12:20] <linyos> here's another possibility. have the network make lots of money for the rich and politically connected
[12:20] <toad_> it might be judicially decided anyway... there's a supreme court case going through
[12:20] <linyos> so they will go break balls when it's interfered with and starts eating up their bottom lines
[12:20] <toad_> linyos: umm, if it could do that, they would have already built it
[12:20] <linyos> who's to say
[12:20] <linyos> it's a possibility though. get special interests fighting on your side
[12:21] <linyos> money especially
[12:21] <toad_> one possibility is to get ISPs on our side. if they compete on performance, bandwidth caps, and so on, then the ones that provide for filesharers should be in favour of filesharing.
[12:21] <linyos> you've got to have a network that rich people depend on to make lots of money
[12:21] <toad_> like the internet? :)
[12:22] <toad_> you do not need hard anonymity from a network in order to make lots of money out of it
[12:22] <linyos> maybe. isps could jump ship as soon as they find real opposition
[12:22] <toad_> you do not need uncensorability either
[12:22] <toad_> anonymity helps sometimes if you happen to be organized crime
[12:22] <toad_> but that's about it
[12:22] <linyos> of course. the point is to link the privacy and the money-making. so ruining one ruins the other
[12:23] <toad_> the rich and powerful can afford dedicated servers and big pipes to feed them
[12:23] <toad_> the economic majority can't
[12:23] <toad_> unfortunately the economic majority are usually forgotten as we see in the software patents debacle
[12:23] <linyos> the practice remains to be worked out, yeah
[12:24] <linyos> and on second thought, it's pretty hard to make it so government proxies would damage any special interest's welfare
[12:25] <toad_> of course not, the government and the megacorps are for all practical purposes one and the same thing in a fascist state
[12:25] <toad_> which both the US and China run the constant risk of becoming
[12:25] <toad_> now, you can think about this
[12:25] <toad_> and i'm going to get some food
[12:25] <toad_> and when i come back, i'm going to think about more immediate problems
[12:25] <linyos> thinking. i've never done that before...
[12:25] <linyos> later
[12:26] <toad_> i believe that if we solve the harvesting problem, we may be able to build something that in future could be developed into, or inspire, something that was really useful even in really hostile regimes
[12:26] <toad_> but there is some element of faith there
[12:26] <toad_> HOWEVER I'm not prepared to work on something that can be shut down for $500 a year
[12:26] <toad_> with any conceivable steganography
[12:26] <toad_> i think that now is the right time to address the harvesting problem
[12:27] <toad_> that's a rough summary of my position in response to what thelema said
[12:27] <toad_> also it's possible we've got our priorities upside down and the contributions freenet makes will be more technical than political in the long run
[12:27] <toad_> but anyway
[12:27] <toad_> i go
[12:28] <thelema> automated sharing of node references will have to go away (except on a maybe monthly frequency)
[12:28] <thelema> I need to go too. Kwa Heri.
[12:28] <linyos> you're just waiting for them to push the button. that's all.
[12:28] <linyos> later
[12:28] <toad_> linyos: I think we can do stuff even if they do
[12:28] <toad_> it just won't be real time
[12:28] <toad_> but it might use similar algorithms
[12:29] <toad_> but bbiab
[12:29] <linyos> no way to circumvent a nasty firewall
[12:29] <linyos> it's just impossible
[12:29] <toad_> sure there is
[12:29] * thelema (~thelema@193.220.86.4) Quit ("thelema has no reason")
[12:29] <linyos> stego fails
[12:29] <toad_> you can always tunnel stuff, especially if you don't mind long latencies
[12:29] <toad_> anyway bbl
[12:29] <toad_> ~ 20-60 minutes
[12:47] <linyos> it's so stupid. i worried all that time about how to prevent identification of darknet traffic, when that is not even necessary. you just push the button and lock down the network's front end.
[12:51] <linyos> and once that happens, the only networking you will be doing is by burning CDs and humping them around
[12:51] <greycat> pigeons! :)
[12:51] <linyos> so we can design a nice Freenet brand pouch to put the CDs in
[12:51] <linyos> because that's all we're good for
[12:52] <greycat> linyos: wireless?
[12:52] <linyos> easily suppressed
[12:52] <linyos> wires would be better
[12:52] <linyos> so we just lay a hundred thousand miles of ethernet cable in various directions
[12:52] <linyos> without being noticed
[12:53] <linyos> and freenet can run over thtaaa
[12:53] <linyos> and would you believe i started this day an optimist
[12:54] <linyos> who dared to think that political power could be curtailed with technology
[12:54] <linyos> the answer is probably to be a moron, because then you never realize that you're wrong
[12:55] <linyos> believe all sorts of nice things that aren't true
[12:58] * sanity (~ian@81-178-91-152.dsl.pipex.com) Quit (Remote closed the connection)
[12:59] * sanity (~ian@81-178-91-152.dsl.pipex.com) has joined #freenet
[13:01] <linyos> i suppose one bit of power you do have is this: you can put together your new private network and force the ruler to decide whether to push the button or not
[13:01] <linyos> so he'd push it and you're worse off than before
[13:17] * goatee (~goatee@ip216-239-81-189.vif.net) Quit (tolkien.freenode.net irc.freenode.net)
[13:18] * goatee (~goatee@ip216-239-81-189.vif.net) has joined #freenet
[13:31] <linyos> hey, i made an image that expresses my wonderful pleasant mood this morning
[13:31] <linyos> http://znex.org/freenet10.jpg
[13:31] <linyos> the future of freenet, d00d
[14:03] * nextime (~nextime@213-140-6-96.fastres.net) Quit ("nextime has no reason")
[14:13] <toad_> wireless is not easily suppressed
[14:13] <toad_> not unless it is actually illegal
[14:13] * moskau23 (~Miranda@dsl-082-082-234-110.arcor-ip.net) has joined #freenet
[14:13] <toad_> and that would be rather obvious
[14:13] <toad_> significant cost in economy and mindshare
[14:14] <toad_> and even then there may be options
[14:15] <toad_> especially with cheap deployable MIMO kit
[14:20] <toad_> on the internet, there is email. there is video-conferencing and VoIP. there are webcams and personal websites unless you make those illegal (nonzero economic and publicity cost).
[14:20] <toad_> there are loads of options and it's way too early to give up
[14:21] * gregh (~greg@defender.hewgill.net) Quit (Read error: 110 (Connection timed out))
[14:21] <toad_> and if the state was so desperate that they actually banned hardware wireless kit, then THAT IS PRETTY DESPERATE!
[14:21] <toad_> even if they were to require a license for it that would be a significant cost
[14:22] <toad_> while it is theoretically possible for them to circumvent or prohibit almost any measure, it does become increasingly expensive
[14:24] * MrNaughty (MrNaughty@d199-126-25-30.abhsia.telus.net) Quit ("\(^_^)/' No Soliciting!!! Unless you have legs way, way up and really, really big tits....")
[14:59] * nextgens (~nextgens@jabber.hst.ru) has left #freenet
[15:10] <linyos> and all those internet services can be locked down
[15:10] <linyos> but i am too sleepy to feel bad about today's failure
[15:10] <toad_> they can be prohibited outright
[15:11] <toad_> but that would be a really bad idea
[15:11] * m[a]zzanet (~irc@mazzanet.user) has joined #freenet
[15:11] <toad_> if they are not prohibited outright then they can be used by us
[15:11] <toad_> bbiab
[15:11] * linyos contemplates sleeping
[15:14] * moskau23 (~Miranda@dsl-082-082-234-110.arcor-ip.net) Quit ("Miranda IM! Smaller, Faster, Easier. http://miranda-im.org")
[15:17] * moskau23 (~Miranda@dsl-082-082-234-110.arcor-ip.net) has joined #freenet
[15:25] * gregh (~greg@defender.hewgill.net) has joined #freenet
[15:25] * mazzanet (~irc@mazzanet.user) Quit (Read error: 110 (Connection timed out))
[15:25] * m[a]zzanet is now known as mazzanet
[15:31] * nextime (~nextime@213-140-6-96.fastres.net) has joined #freenet
[15:58] * gregh_ (~greg@defender.hewgill.net) has joined #freenet
[15:58] * gregh (~greg@defender.hewgill.net) Quit (Read error: 104 (Connection reset by peer))
[16:03] <linyos> ????????? ?????????.
[16:03] * linyos (~asdfasdf@adsl-68-79-4-86.dsl.emhril.ameritech.net) has left #freenet
[16:05] <toad_> serious problems on dad's computer :<
[16:05] <greycat> he wasn't supposed to sleep on the keyboard.
[16:05] * toad_ downloading knoppix so can extract as much files as poss.
[16:06] <toad_> i suppose i could put the disk into servalan, but i think that would be unwise
[16:06] * gregh_ is now known as gregh
[16:26] * greycat (rfc1413@wooledge.org) Quit ("This time the bullet cold rocked ya / A yellow ribbon instead of a swastika")
[16:34] * FallingBuzzar1 (~srademach@66.151.22.70) has joined #freenet
[16:34] * FallingBuzzard (~srademach@66.151.22.70) Quit (Read error: 104 (Connection reset by peer))
[17:06] * Romster (Romster@203.129.145.209) Quit (Read error: 60 (Operation timed out))
[17:27] * Romster (~Romster@203.129.154.22) has joined #freenet
[17:57] * FallingBuzzar1 (~srademach@66.151.22.70) has left #freenet
[18:56] * cbreak (~cbreak@84-72-98-199.dclient.hispeed.ch) Quit ("leaving")
[19:00] * Hory (~Miranda@82.78.27.85) Quit (Read error: 60 (Operation timed out))
[19:02] * Hory (~Miranda@82.78.27.85) has joined #FreeNet
[20:31] * sanity (~ian@81-178-91-152.dsl.pipex.com) Quit ()
[20:33] * moskau23 (~Miranda@dsl-082-082-234-110.arcor-ip.net) Quit (Read error: 145 (Connection timed out))
[22:34] * Romster1 (~Romster@203.129.153.68) has joined #freenet
[22:41] * goatee (~goatee@ip216-239-81-189.vif.net) Quit (Read error: 104 (Connection reset by peer))
[22:44] * TheSeeker (Fridlekh@pool-71-107-148-131.lsanca.dsl-w.verizon.net) has joined #freenet
[22:56] * Hory (~Miranda@82.78.27.85) Quit ("CyberLore.net - Recommendations on the best games, freeware and websites.")
[22:56] * Romster (~Romster@203.129.154.22) Quit (Read error: 110 (Connection timed out))
[23:13] * goatee (~goatee@ip216-239-81-141.vif.net) has joined #freenet
These logs were automatically created by Jay Oliveri with his gimp hapi on irc.freenode.net.
