i2p.www/pages/meeting60.html

<pre>

[22:53] &lt;jrand0m&gt; 0.x) welcome 
[22:53] &lt;jrand0m&gt; 1.x) todo before i2p 0.2: 
[22:53] &lt;jrand0m&gt; 1.1) bw limiting 
[22:54] &lt;jrand0m&gt; 1.2) AES the tcp transport 
[22:54] &lt;jrand0m&gt; 1.3) package up [distro, building, seeding, running] 
[22:54] &lt;jrand0m&gt; 1.4) ElG session ack discussion 
[22:54] * hezekiah is away: Is off eating to keep up is strength for his upcoming tests
[22:54] &lt;jrand0m&gt; 2.x) apps 
[22:54] &lt;jrand0m&gt; 2.1) ns 
[22:54] &lt;jrand0m&gt; 2.2) i2pim 
[22:54] &lt;jrand0m&gt; 2.3) i2ptunnel 
[22:54] &lt;jrand0m&gt; 3.x) ??? 
[22:54] &lt;jrand0m&gt; 0.x) welcome 
[22:54] &lt;jrand0m&gt; welcome to meeting 60 
[22:54] &lt;dm&gt; thanks
[22:55] &lt;mihi&gt; oh. 60. whe have something to celebrate ;)
[22:55] * jrand0m hands out some beers
[22:55] *** wiht (anon@anon.iip) has joined channel #iip-dev
[22:55] * mihi does not like alcohol
[22:55] &lt;thecrypto&gt; hi
[22:55] * jrand0m takes back the german beer
[22:55] &lt;thecrypto&gt; i'm not awake
[22:55] &lt;mihi&gt; but not asleep either ;)
[22:55] &lt;jrand0m&gt; heh interesting 
[22:56] &lt;jrand0m&gt; ok, 1.{x,1}) todo before 0.2 
[22:56] &lt;thecrypto&gt; i just remembered the meeting it today
[22:56] &lt;thecrypto&gt; i was just discussing wireless cards
[22:56] &lt;jrand0m&gt; the router is in pretty good shape, and I think 0.2 will be ready for internal dist and testing 
[22:57] *** Signoff: mihi (EOF From client)
[22:57] &lt;jrand0m&gt; there are three things that need to get implemented first though.  bandwidth limiting, AES on the TCP, and ElG session ack 
[22:57] &lt;jrand0m&gt; the bandwidth limiting will be classless for the moment - just one big honking input limit and output limit. 
[22:58] *** mihi_backup (~mihi@anon.iip) has joined channel #iip-dev
[22:58] &lt;jrand0m&gt; the code is committed to tie it in, but the limit is set at infinity atm 
[22:58] &lt;jrand0m&gt; for point 1.2, the TCP transport has key exchange and authentication, but needs encryption 
[22:59] &lt;jrand0m&gt; nop/semi-awake-thecrypto&gt; thoughts on AES for streaming?   
[22:59] &lt;thecrypto&gt; carefully
[22:59] &lt;jrand0m&gt; !thwap 
[22:59] &lt;thecrypto&gt; sorry
[23:00] * CounterRev eats popcorn in the grand stands
[23:00] &lt;jrand0m&gt; CBC using first 16 bytes of H(sessionKey) work, or should we do something else? 
[23:00] &lt;dm&gt; I can't remember where that originated, exclamation mark before an action.
[23:00] &lt;thecrypto&gt; i realized when the agenda started being pasted today was tuesday
[23:00] &lt;jrand0m&gt; dm&gt; some eggdrops use them as command identifiers 
[23:01] &lt;thecrypto&gt; why not use more bits from the DH for that?
[23:01] &lt;jrand0m&gt; hmm? 
[23:01] &lt;thecrypto&gt; i don't see any problem for that
[23:01] &lt;thecrypto&gt; for the IV
[23:01] &lt;jrand0m&gt; more bits of what?  X, Y, x, y? 
[23:02] *** mihi (mihi@anon.iip) has joined channel #iip-dev
[23:02] &lt;thecrypto&gt; K
[23:02] &lt;dm&gt; !thwap thecrypto
[23:02] &lt;thecrypto&gt; what?
[23:02] &lt;dm&gt; he said X,Y,x or y
[23:02] &lt;jrand0m&gt; AES 256 only has a 16 byte K 
[23:02] &lt;jrand0m&gt; er, IV 
[23:03] &lt;jrand0m&gt; the session key is 32 bytes 
[23:03] &lt;thecrypto&gt; but you generate many more
[23:03] &lt;jrand0m&gt; (originally I was stupid and made the IV the first 16 bytes of the key ;) 
[23:03] &lt;thecrypto&gt; so use 48 bytes from the DH
[23:04] &lt;dm&gt; we all make mistakes.
[23:04] &lt;jrand0m&gt; hmm ok, pulling more bytes off the modPow will work.  cool.  but why 48?   
[23:04] &lt;jrand0m&gt; (why not just another 16?) 
[23:05] &lt;thecrypto&gt; i meant 32+16
[23:05] &lt;jrand0m&gt; ah 'k 
[23:06] &lt;jrand0m&gt; ok cool, CBC with that.  if you or nop or someone else has time (userx?), we can get rekeying / rotating keys in there.  but for the moment I'll get it CBC'ed w/ that IV 
[23:06] &lt;jrand0m&gt; (rekey would be just drop the connection and re-DH) 
[23:06] &lt;jrand0m&gt; ok, 1.3) packaging up for 0.2 
[23:07] &lt;jrand0m&gt; that just means a make / ant script, install / build doc, plus a method for distributing seed routerInfo references  
[23:07] &lt;jrand0m&gt; (plus, most likely, a set of a few routers for people to connect to) 
[23:08] &lt;jrand0m&gt; I hope to have 0.2 out and ready for people to d/l and install by this time next week, if not sooner 
[23:09] &lt;jrand0m&gt; the one last point holding up 0.2 is the ElG session ack 
[23:09] &lt;dm&gt; can we use 0.2 to talk to each other?
[23:09] &lt;jrand0m&gt; yes dm 
[23:09] &lt;dm&gt; ok, that's useful.
[23:10] &lt;mihi&gt; dm: is talking with you really useful? *veg*
[23:10] &lt;dm&gt; (Killer App: Cyber Sex)
[23:10] &lt;jrand0m&gt; 0.2 is definitely not even alpha, so it won't be for the faint of heart 
[23:10] &lt;jrand0m&gt; heh 
[23:10] &lt;jrand0m&gt; (but we need to get some testing and experience on some platforms and usage patterns other than my xp box here) 
[23:11] &lt;dm&gt; I'll take care of the marketing for I2P, I have a knack for it.
[23:11] &lt;jrand0m&gt; lol 
[23:11] &lt;dm&gt; 0.2 code name: CyberSex, features, anonymous CyberSex through text.
[23:11] &lt;jrand0m&gt; ok, the ElG session ack stuff.   
[23:11] &lt;wiht&gt; I can test on FreeBSD and Gentoo Linux.
[23:11] &lt;jrand0m&gt; awesome wiht! 
[23:11] * popopopo can test Debian
[23:11] &lt;jrand0m&gt; nice 
[23:12] &lt;thecrypto&gt; i can test on mandrake
[23:12] * dm can test windows XP
[23:12] &lt;jrand0m&gt; damn we're going to have pretty good coverage :) 
[23:12] * dm stands out from the crowd.
[23:12] &lt;popopopo&gt; I can test win2k as well I think, although that should be about the same as XP I think
[23:12] &lt;jrand0m&gt; hopefully 
[23:12] &lt;dm&gt; stick to unix popopopo
[23:13] * thecrypto plays the trump: I can test win 98 :)
[23:13] &lt;jrand0m&gt; uuuugh 
[23:13] * dm welcomes his little brother to the crowd.
[23:13] &lt;popopopo&gt; I have an old DOS disk around here. Will java run on that? :)
[23:13] &lt;wiht&gt; popopopo: Probably not.
[23:13] &lt;jrand0m&gt; actually, probably. 
[23:14] &lt;jrand0m&gt; since gcc supports java 
[23:14] &lt;jrand0m&gt; and djgpp runs on dos 
[23:14] &lt;popopopo&gt; I2P on DOS, now that would be an accomplishment
[23:14] &lt;mihi_backup&gt; popopopo: there is a commercial java version for DOS ;)
[23:14] * jrand0m is currently hacking around with gjc to see if we can compile to native
[23:14] &lt;dm&gt; Getting PPP working on DOS would be accomplishment enough.
[23:14] &lt;mihi&gt; dm: use arachne.
[23:15] &lt;popopopo&gt; gcj
[23:15] &lt;mihi&gt; should work via parallel modem.
[23:15] &lt;nop&gt; oh shit
[23:15] &lt;nop&gt; it's tuesday
[23:15] &lt;nop&gt; haha
[23:15] &lt;jrand0m&gt; ah right popopopo... I always type it wrong 
[23:15] &lt;mihi&gt; oops, serial modem ;)
[23:15] &lt;jrand0m&gt; wb nop ;) 
[23:15] &lt;thecrypto&gt; nop: you and me both
[23:15] &lt;nop&gt; haha
[23:16] &lt;nop&gt; sorry
[23:16] &lt;nop&gt; my brain is fried this week
[23:16] &lt;nop&gt; re-cooping
[23:16] &lt;nop&gt; from this weekend
[23:16] &lt;thecrypto&gt; nop: you and me both
[23:16] &lt;jrand0m&gt; ok, the ElG ack stuff. 
[23:16] &lt;jrand0m&gt; (take 4) 
[23:16] * CounterRev could test xp as well "its running.. its running... its stopped"
[23:17] &lt;jrand0m&gt; right now the ElG+AES works by using session tags - 32 byte tags prepended to AES streams to identify what session key to decrypt with 
[23:17] &lt;mihi&gt; and "it's bluescreened, it's autorebooting"
[23:17] &lt;jrand0m&gt; (these tags are distributed in the streams, and the session keys are distributed in the ElG) 
[23:18] &lt;jrand0m&gt; the problem comes up if we lose the first ElG+AES, we've lost the session key as well as the initial set of session tags 
[23:18] &lt;thecrypto&gt; brb, playing magic
[23:18] &lt;jrand0m&gt; so we need to make sure messages use ElG instead of AES with a session tag unless we know for sure the session key has been received 
[23:19] &lt;jrand0m&gt; there are lots of optimizations and tradeoffs here - one could always ElG and just forget about acks 
[23:19] &lt;jrand0m&gt; (trading cpu for bandwidth) 
[23:19] &lt;jrand0m&gt; or one could always require an ack for the first block, and then AES 
[23:20] &lt;jrand0m&gt; but that breaks if the recipient crashes and loses the session key 
[23:20] * nop is backlogging
[23:20] &lt;jrand0m&gt; there are a few different patterns that can be used to implement these sessions 
[23:21] &lt;jrand0m&gt; what I'm probably going to go with is requiring an ack for the ElG message, using AES from then on, but every {$timeperiod, numMessages, bytes} send another ElG 
[23:22] &lt;dm&gt; like P frames in MPEG!
[23:22] &lt;jrand0m&gt; 'k, I dont know the MPEG standard really 
[23:22] &lt;dm&gt; sorry, I frames.
[23:23] &lt;jrand0m&gt; the thing is, these ElG messages are expensive.  ~0.5-1.0 seconds 
[23:23] &lt;jrand0m&gt; (cpu time) 
[23:23] &lt;jrand0m&gt; though, hmm, we could precalculate them 
[23:23] &lt;jrand0m&gt; (if we don't rotate session keys) 
[23:23] &lt;mihi&gt; which cpu? ;)
[23:23] &lt;dm&gt; sender or reciever cpu time?
[23:24] &lt;jrand0m&gt; exactly mihi - mine.  p4 1.7g.  so we need some perf tuning on the algorithm 
[23:24] &lt;jrand0m&gt; both dm 
[23:24] &lt;jrand0m&gt; (.5 encrypt, .5 decrypt) 
[23:24] &lt;jrand0m&gt; (hand wavey approx) 
[23:24] * mihi has celeron 700
[23:24] * dm has PIII 850
[23:24] &lt;jrand0m&gt; cool.  I'm also getting a sun ultra1 (150Mhz) in a few weeks 
[23:25] &lt;dm&gt; Why you getting a SUN?
[23:25] &lt;jrand0m&gt; I'm leaving the option open for reevaluating our ElG keysize if we need to 
[23:26] &lt;jrand0m&gt; dm&gt; its free (you can buy a u1 for ~50-100 USD).  I also have a few sparc for various purposes 
[23:26] &lt;dm&gt; sweet.
[23:27] &lt;jrand0m&gt; ok, unless anyone has any comments on the ElG stuff, on to agenda item 2.x... 
[23:27] &lt;jrand0m&gt; ok, 2.x) apps 
[23:27] &lt;jrand0m&gt; 2.1) ns 
[23:27] &lt;jrand0m&gt; wiht, how goes? 
[23:28] *** Signoff: mihi_backup (Ping timeout)
[23:30] &lt;wiht&gt; I have not done any coding on the naming server this past week.
[23:30] &lt;wiht&gt; I should say "of" the naming server.
[23:31] &lt;jrand0m&gt; 'k, understandable.. what do you think your schedule looks like for the next few weeks for implementation?  no committment or anything, just wondering 
[23:31] &lt;wiht&gt; But I will be working on it this week, and should have something more to report during the next meeting.
[23:31] &lt;jrand0m&gt; cool, great 
[23:31] &lt;CounterRev&gt; what is i2p like sans ns?
[23:32] &lt;jrand0m&gt; icq, I suppose 
[23:32] &lt;wiht&gt; I should be less busy during the next few weeks, and will have time for this. I cannot say when the naming server will be completed at this time.
[23:32] &lt;jrand0m&gt; you need the secret number for contacting someone, but once you have it, you're ok 
[23:33] &lt;mihi&gt; CounterRev: lots of very long destination keys...
[23:33] &lt;jrand0m&gt; CounterRev&gt; I suspect someone will install a squid and an i2ptunnel on their router, allowing people to point at the destination and browse the web anonymously 
[23:33] *** mihi_backup (~mihi@anon.iip) has joined channel #iip-dev
[23:33] &lt;jrand0m&gt; ok, 2.2) i2pim... 
[23:33] &lt;jrand0m&gt; thecrypto&gt; you still playing magic?   
[23:34] &lt;dm&gt; i2pim, is that the name of the IM app?
[23:34] &lt;jrand0m&gt; yup 
[23:34] &lt;wiht&gt; dm: Yes.
[23:34] &lt;dm&gt; crap! i2m, Invisible Instant messaging.
[23:34] &lt;dm&gt; i^2m, just like i^2p!
[23:34] &lt;jrand0m&gt; sounds like there's a market 
[23:35] &lt;mihi&gt; dm: we had that discussion last meeting...
[23:35] &lt;dm&gt; fine fine... I'll have to live with it.
[23:36] &lt;jrand0m&gt; ok, thecrypto is probably still off.  he'll post to the list if there's any news on the i2pim 
[23:36] &lt;jrand0m&gt; (i hope) 
[23:36] &lt;jrand0m&gt; ok, 2.3) i2ptunnel 
[23:36] * jrand0m waves to mihi
[23:37] &lt;wiht&gt; jrand0m: I see that we have three applications being developed at this time. Have others been proposed?
[23:37] &lt;mihi&gt; oops...
[23:37] &lt;mihi&gt; but i have nothing to tell anyway ;)
[23:37] &lt;mihi&gt; i2p is in cvs; look @ it ;)
[23:37] &lt;dm&gt; mihi speaks in code.
[23:38] &lt;jrand0m&gt; heh word.  any updates since you added that config&gt; command? 
[23:38] &lt;dm&gt; or rather, his code speaks for itself.
[23:38] &lt;mihi&gt; i2ptunnel is in cvs i mean...
[23:38] * nop proposes ex-lax for i2p
[23:38] &lt;mihi&gt; jrand0m: nope.
[23:38] &lt;jrand0m&gt; coo' 
[23:38] &lt;jrand0m&gt; would that be like a LaTeX clone nop? 
[23:39] &lt;jrand0m&gt; wiht&gt; some various discussions, but those three have been the main ones discussed 
[23:39] &lt;jrand0m&gt; (the i2ptunnel enables a shitload of normal apps to operate over it) 
[23:40] &lt;dm&gt; has it been tested with anything?
[23:40] &lt;mihi&gt; i2ptunnel has been tested extensively with the local router.
[23:40] &lt;dm&gt; (not locally)
[23:40] &lt;jrand0m&gt; I've done some web browsing with it, as well as ssh 
[23:40] &lt;jrand0m&gt; (browsing a single site, that is) 
[23:41] * wiht wonders if that website can still track people through cookies.
[23:41] &lt;dm&gt; great.
[23:41] &lt;jrand0m&gt; sure wiht, cookies are still sent  
[23:43] &lt;jrand0m&gt; ok, thats it for apps atm, and on to 3.x) ??? 
[23:43] &lt;jrand0m&gt; any questions, thoughts, concerns, toenail clippers, hedgehogs? 
[23:43] &lt;mihi&gt; a pony? ;)
[23:43] &lt;jrand0m&gt; yeah, I want a pony! 
[23:43] * dm raises his glass.
[23:43] &lt;mihi&gt; nop will give you onw...
[23:43] &lt;dm&gt; TO 0.2!!!
[23:44] &lt;jrand0m&gt; oh yeah, mihi pointed out that in 17 minutes we miss the deadline I set in my flog a month ago 
[23:44] &lt;jrand0m&gt; to 1.0! 
[23:45] &lt;mihi&gt; 7 minutes ;)
[23:45] &lt;dm&gt; so 2 weeks from now, we'll have 20 people on the network who'll be able to send texts to each other?
[23:45] &lt;jrand0m&gt; I said we'd have an alpha out by end of september.  well, it looks like we'll be a little late on that, but I think we're still making a good pace 
[23:45] &lt;jrand0m&gt; yes dm (and files, etc) 
[23:45] &lt;nop&gt; sickening
[23:45] &lt;jrand0m&gt; there will still be room for performance improvements, of course.  there hasn't been much tuning at all 
[23:45] &lt;nop&gt; sorry, I already spent my budget the next month
[23:45] &lt;dm&gt; Alrighty, well if that happens, I will applaud.
[23:45] &lt;nop&gt; pony has to wait
[23:46] &lt;jrand0m&gt; damnit! 
[23:46] &lt;nop&gt; bought a grand piano instead
[23:46] &lt;nop&gt; ;)
[23:46] &lt;jrand0m&gt; heh nice 
[23:47] &lt;mihi&gt; jrand0m: here are some virtual ponies for you: (Link: http://tinyurl.com/p8kx)http://tinyurl.com/p8kx
[23:47] &lt;dm&gt; estimated latency?
[23:47] &lt;dm&gt; with network &lt; 30 nodes?
[23:48] &lt;jrand0m&gt; dm&gt; depends on the sender and receiver's tunnel length 
[23:48] &lt;dm&gt; order of magnitude.
[23:48] &lt;jrand0m&gt; (e.g. if they have 0 length tunnels, it'll be approx 400ms) 
[23:49] &lt;dm&gt; 0 length tunnels, don't know how I2P works, but 0 length doesn't sound good ;)
[23:49] &lt;mihi&gt; or (Link: http://tinyurl.com/p8l7)http://tinyurl.com/p8l7
[23:49] &lt;jrand0m&gt; if they have 2 hop tunnels, the latency should be ~ 1s 
[23:49] &lt;dm&gt; alrighty.
[23:49] &lt;dm&gt; throughput?
[23:50] &lt;dm&gt; harder to estimate?
[23:50] &lt;jrand0m&gt; depends on bandwidth 
[23:51] &lt;dm&gt; Say X is the non I2P bandwidth between 2 nodes, through all the hops I2P uses.
[23:51] &lt;dm&gt; what is the equation for Y(X) where Y is the throughput of I2P from one node to the other.
[23:51] &lt;jrand0m&gt; I2P doesn't add a significant overhead to the size of the data 
[23:52] &lt;dm&gt; alrighty.
[23:52] &lt;dm&gt; wait and see I guess.
[23:53] &lt;jrand0m&gt; once she's up and running, we'll definitely get some modeling going on 
[23:53] &lt;dm&gt; aye..
[23:54] * dm raises his glass.
[23:54] &lt;dm&gt; To women!
[23:54] &lt;mihi&gt; dm: you forgot to lower it first ;)
[23:55] &lt;dm&gt; You germans...
[23:55] &lt;dm&gt; You wacky germans.
[23:55] &lt;jrand0m&gt; ok, I think thazzabout it 
[23:55] &lt;dm&gt; It's no suprise an american is leading us into the 21st century of anonymous networking.
[23:55] * mihi /igs dm
[23:55] &lt;jrand0m&gt; american? 
[23:56] &lt;mihi&gt; jrand0m is spanish i guess.
[23:56] &lt;dm&gt; I was referring to you!
[23:56] &lt;mihi&gt; or italian or sth in CET timezone.
[23:56] &lt;dm&gt; jrand0m, the quiet american.
[23:56] &lt;wiht&gt; dm: jrand0m seems to be a European.
[23:56] &lt;dm&gt; !thwap mihi
[23:57] &lt;dm&gt; !thwap wiht
[23:57] &lt;dm&gt; jrand0m: your team members don't know you well.
[23:58] * jrand0m thinks dm has been drinkin a lil too much, but its time for me to do the same. 
[23:58] * jrand0m raises the *baf*...
[23:58] &lt;dm&gt; nice try jr!
[23:58] * jrand0m *baf*s the meeting away
</pre>