Files
i2p.www/pages/meeting49.html
duck 50712ca70f "Screw you guys, I am going to do it myself."
add missing meetings and how does it work pages
(duck)
2004-07-15 20:11:59 +00:00

743 lines
37 KiB
HTML

<pre>
--- Log opened Tue Jul 15 17:46:47 2003
17:46 &lt; gott&gt; yo.
17:46 &lt;@nop&gt; just a heads up on my silence
17:46 &lt;@hezekiah&gt; Tue Jul 15 21:46:49 UTC 2003
17:47 &lt;@hezekiah&gt; OK. The iip-dev meeting has started.
17:47 &lt;@hezekiah&gt; Is it the 48th or 49th?
17:47 &lt; jrand0m&gt; nop&gt; this is why its critical that we get the router
architecture pounded out asap. I understand that different people have
different rates of speed, and we must segment so different components can
proceed accordingly
17:47 &lt; mihi&gt; 49th
17:47 &lt;@hezekiah&gt; OK! Welcome to the 49th iip-dev meeting!
17:47 &lt; jrand0m&gt; I have three more days at my job, after which 90+ hours /
week will be dedicated to getting this going
17:48 &lt; jrand0m&gt; I know and don't expect everyone to be able to do that,
which is why we need to segment
17:48 &lt; jrand0m&gt; hi hezekiah :)
17:48 &lt;@hezekiah&gt; lol
17:48 &lt;@nop&gt; to rebutt on that
17:48 &lt;@hezekiah&gt; I'll wait a minute. Then we can do the agenda. :)
17:48 &lt;@nop&gt; the security of the router architecture is dependant that you
do not rush as well
17:49 &lt;@nop&gt; if we do
17:49 &lt;@nop&gt; we overlook
17:49 &lt;@nop&gt; which could leave us cleaning up a big mess later
17:49 -!- Rain [Rain@anon.iip] has quit [I Quit]
17:49 &lt; jrand0m&gt; nop&gt; disagree. we can still build app layer and APIs
without implementing the router (or even knowing how the network will operate)
17:49 &lt;@nop&gt; I agree with that
17:50 &lt;@nop&gt; I'm specifically talking about the underlying network
17:50 &lt; jrand0m&gt; if we can agree to the API I sent out, then thats the
segmentation we need
17:50 &lt; jrand0m&gt; right, router impl and network design still isn't done
17:50 &lt;@nop&gt; ok
17:50 &lt;@nop&gt; oh, I can definitely agree with your api so far
17:51 &lt;@hezekiah&gt; jrand0m: One problem.
17:51 &lt; jrand0m&gt; shoot hezekiah
17:51 &lt;@hezekiah&gt; It will look different if you implement it in C.
17:51 &lt; jrand0m&gt; not too different
17:51 &lt; gott&gt; oh dear
17:51 &lt; jrand0m&gt; less capital letters, and replace the objects with structs
17:51 &lt; gott&gt; what languages are people considering implementing it in?
17:51 &lt; jrand0m&gt; (for the api)
17:51 &lt;@hezekiah&gt; Uh, jrand0m? There is no 'byte[]' in C.
17:51 &lt; jrand0m&gt; gott&gt; read the mail archives for some example answers to that
17:52 &lt;@hezekiah&gt; You will be using void*'s with an integer to specifiy the
length most likely.
17:52 &lt; jrand0m&gt; hezekiah&gt; then unsigned int[]
17:52 &lt; gott&gt; jrand0m: for once, a religious war that I'm not a part of
17:52 &lt;@hezekiah&gt; If I remember correctly (help me out here nop), you can't
just return an unsigned int[] from a function.
17:53 &lt;@hezekiah&gt; gott: It's not a religious war. I'm just saying that the
API as a concept might be fine, but in C it would look seriously different.
17:53 &lt; gott&gt; hezekiah: as opposed to what? pseudocode?
17:53 &lt; jrand0m&gt; right, syntactic changes. but yes, if there are real
differences, we need to get them worked out ASAP. (like, today) Perhaps
now would be a good tiem to look at the email I sent entitled "high level
router architecture and API" and review?
17:54 &lt;@hezekiah&gt; nop? UserX? Are you game for that?
17:54 &lt; jrand0m&gt; not too different, but different none the less, yes.
which is why I said Java API on todays email :)
17:54 -!- WinBear [WinBear@anon.iip] has joined #iip-dev
17:55 &lt;@nop&gt; wait
17:55 &lt;@nop&gt; reading above
17:55 -!- mihi_2 [~none@anon.iip] has joined #iip-dev
17:55 -!- mihi is now known as nickthief60234
17:55 -!- mihi_2 is now known as mihi
17:55 &lt; jrand0m&gt; wb mihi
17:55 &lt; gott&gt; btw, is this being live logged?
17:55 -!- nickthief60234 [~none@anon.iip] has quit [EOF From client]
17:55 &lt;@hezekiah&gt; gott: Yes.
17:55 &lt; mihi&gt; redundancy rules ;)
17:55 &lt; gott&gt; I'll just read it later on then.
17:55 -!- gott [~gott@anon.iip] has left #iip-dev [gott]
17:56 &lt;@nop&gt; ok
17:56 &lt;@nop&gt; yes
17:56 &lt; WinBear&gt; jrand0m: hi
17:56 &lt;@nop&gt; definitely differences
17:56 &lt;@nop&gt; what we need
17:56 &lt; jrand0m&gt; heya WinBear
17:56 &lt;@nop&gt; is a team of certain developers to write the main api level
controls for these languages
17:56 &lt;@nop&gt; we know that jrand0m can handle java
17:56 &lt;@nop&gt; and probably could team up with thecrypto as well
17:56 &lt;@nop&gt; and hezekiah and the gang can do C
17:56 &lt;@nop&gt; and jeremiah if he's willing
17:56 &lt;@nop&gt; can do python
17:56 &lt;@hezekiah&gt; I can do C++ too! ;-)
17:56 &lt;@nop&gt; ok
17:56 &lt;@nop&gt; C++ as well
17:57 &lt;@hezekiah&gt; lol
17:57 &lt;@nop&gt; C++ will probably work
17:57 &lt;@nop&gt; with C
17:57 &lt;@nop&gt; if you don't template the crap out of it
17:57 &lt; jrand0m&gt; heh
17:57 &lt;@hezekiah&gt; lol
17:57 &lt;@hezekiah&gt; Actually, while MSVC can link C and C++ object files,
gcc doesn't seem to like that.
17:57 &lt;@nop&gt; aka, stick to structs that are compatible with C, or is that
not viable
17:57 &lt; jrand0m&gt; first question, prior to that, is what applications will use
these APIs? I know of apps that will want to use java, will iproxy be in C?
17:58 &lt;@hezekiah&gt; nop: I don't think C and C++ are object compatible.
17:58 &lt;@nop&gt; ok
17:58 &lt;@hezekiah&gt; nop: C++ won't get along with C much better than Java.
17:58 &lt;@nop&gt; well maybe USerX could do C
17:58 &lt;@nop&gt; and you could pull C++
17:58 &lt;@hezekiah&gt; We don
17:58 &lt;@nop&gt; ?
17:58 &lt;@hezekiah&gt; don't even need to _do_ C++ if you don't want to. It's
just that I prefer it.
17:59 &lt;@nop&gt; well, the thing is
17:59 &lt;@nop&gt; there are a lot of C++ developers
17:59 &lt;@nop&gt; especially in the microsoft world
17:59 &lt;@hezekiah&gt; Even in the Linux world. (see: KDE and Qt.)
17:59 &lt; jrand0m&gt; C and C++ are binary compatible if you just make .so or .a
17:59 &lt; jrand0m&gt; (btw)
18:00 &lt;@nop&gt; can C be a good placement for C++, aka C++ developers would be
able to handle a c api easier than a C++ api with a c developer?
18:00 &lt;@hezekiah&gt; jrand0m: Yeah. You can probably have libraries ... but if
you can
18:00 &lt;@hezekiah&gt; jrand0m: can't even use classes, it sorta defeats the
purpose.
18:00 &lt;@nop&gt; right
18:00 &lt;@nop&gt; let's stick with C
18:01 &lt;@nop&gt; because C++ coders can still call a C library rather easily
18:01 &lt;@hezekiah&gt; If one module needs to call anothers functions, then they
had best both be the same language.
18:01 &lt;@hezekiah&gt; nop: C++ coders will know C well enough ... though it
might take some work if they never /learned/ C.
18:02 &lt;@hezekiah&gt; However, C coders wouldn't know C++ since C is just a
subset of C++.
18:02 -!- logger_ [~logger@anon.iip] has joined #iip-dev
18:02 -!- Topic for #iip-dev: logfiles will be online after the meeting:
http://wiki.invisiblenet.net/?Meetings
18:02 [Users #iip-dev]
18:02 [@hezekiah] [+Ehud ] [ leenookx] [ moltar] [ tek ]
18:02 [@nop ] [ jeremiah] [ logger_ ] [ Neo ] [ WinBear]
18:02 [@UserX ] [ jrand0m ] [ mihi ] [ ptsc ]
18:02 -!- Irssi: #iip-dev: Total of 14 nicks [3 ops, 0 halfops, 1 voices,
10 normal]
18:02 &lt; jrand0m&gt; right
18:02 -!- Irssi: Join to #iip-dev was synced in 9 secs
18:02 &lt; jrand0m&gt; (with JMS :)
18:02 &lt;@nop&gt; yep
18:03 -!- You're now known as logger
18:03 &lt; jrand0m&gt; ok, can we review the overall architecture to see whether
the APIs are even relevent first?
18:03 &lt;@nop&gt; fine 18:04 &lt; jrand0m&gt; :)
18:04 &lt; jrand0m&gt; ok, see the email I sent w/ the routerArchitecture.png.
any thoughts on that seperation?
18:04 -!- tek [~tek@anon.iip] has quit []
18:05 &lt; WinBear&gt; jrand0m: is that on the wiki?
18:05 &lt; jrand0m&gt; WinBear&gt; no, on the mailing list, though the archives
are down. lemmie add it to the wikki
18:06 &lt;@hezekiah&gt; Correct me if I'm wrong ...
18:07 &lt;@hezekiah&gt; ... but it looks like we're going to have 3 seperate API's
that are as similar as possible.
18:07 &lt;@hezekiah&gt; Right?
18:07 &lt; jrand0m&gt; yes hezekiah
18:07 &lt;@hezekiah&gt; So since each API is in a different language, are they
going all each have seperate implementations?
18:07 &lt; jrand0m&gt; yes
18:07 &lt;@hezekiah&gt; Or is there a way for Java or Python to access a C library?
18:08 &lt; jrand0m&gt; yes, but we don't want to go that route
18:08 &lt; mihi&gt; for java: JNI
18:08 &lt;@hezekiah&gt; So this talk about Java, C, C++, Python, etc. working
together is mute since they never will?
18:08 &lt; jrand0m&gt; how do I attach an image to the wiki?
18:08 &lt;@hezekiah&gt; Each API has its own backend written in that language.
18:08 &lt; jrand0m&gt; no hezekiah, look at the diagram
18:09 &lt;@hezekiah&gt; Oh, duh!
18:09 &lt;@hezekiah&gt; The API's don't link to a backend.
18:10 &lt;@hezekiah&gt; They talk via sockets.
18:10 &lt; jrand0m&gt; si sr
18:10 &lt;@hezekiah&gt; This is still a little confusing though.
18:10 &lt;@hezekiah&gt; Give me a sec here. :)
18:11 &lt;@hezekiah&gt; OK. What is the thing labeled 'transport'?
18:11 &lt; jrand0m&gt; for example, bidirectional HTTP transport, SMTP transport,
plain socket transport, polling HTTP socket, etc
18:11 &lt; jrand0m&gt; the thing that moves bytes between routers
18:12 &lt;@hezekiah&gt; OK.
18:12 &lt;@hezekiah&gt; So the diagram I'm looking at shows one person's computer.
18:12 &lt;@hezekiah&gt; He has a router that talks to other people's computers
via the transports.
18:12 &lt; jrand0m&gt; correct
18:12 &lt;@hezekiah&gt; Person 1 (Alice) has 2 applications running.
18:12 &lt;@hezekiah&gt; One is in C, the other in Java.
18:13 &lt;@hezekiah&gt; Both are linked to a library (that's the API).
18:13 &lt; jrand0m&gt; both are "linked" to seperate libraries (the APIs)
18:13 &lt;@nop&gt; simple concept
18:13 &lt;@nop&gt; yes
18:13 &lt;@hezekiah&gt; Those libraries, take input from the program encrypt it,
and send it via sockets (unix or TCP) to the router ... which is another
program Alice is running.
18:13 &lt; jrand0m&gt; correct
18:14 &lt;@hezekiah&gt; OK. So it's kinda like isproxy being split in two.
18:14 &lt; jrand0m&gt; bingo :)
18:14 &lt;@hezekiah&gt; One part is low end and written in C, and the other is
high end and written in whatever.
18:14 &lt; jrand0m&gt; exactly
18:14 &lt;@hezekiah&gt; OK. I get it. :)
18:14 &lt; jrand0m&gt; w00t
18:14 &lt;@hezekiah&gt; So no language needs to play nice with any other language.
18:14 &lt; jrand0m&gt; WinBear&gt; sorry, I can't toss it on the wiki as it only
takes text :/
18:15 &lt;@hezekiah&gt; Since they all comunicate with the router via sockets,
you could write an API in PASCAL for all the design cares.
18:15 &lt;@nop&gt; yes
18:15 &lt;@nop&gt; arbitrary
18:15 &lt; jrand0m&gt; right
18:15 &lt;@nop&gt; it handles arbitrary sockets
18:15 &lt; jrand0m&gt; though some things need to be standardized (like the data
structures for Destination, Lease, etc)
18:15 &lt; WinBear&gt; jrand0m: i get a vague idea based on what hezekiah is saying
18:15 &lt; jrand0m&gt; word
18:16 &lt;@hezekiah&gt; jrand0m: Right. The structure and order of the bytes that
go across that socket is set in a design somewhre
18:16 &lt;@hezekiah&gt; somewhere.
18:17 &lt;@hezekiah&gt; But you can still implement how those bytes are send and
received any joly way you please.
18:17 &lt;@nop&gt; WinBear: it's the same exact way that the irc client works
with isproxy
18:17 &lt; jrand0m&gt; exactly
18:17 &lt;@hezekiah&gt; Good.
18:17 &lt;@hezekiah&gt; I understand now. :)
18:17 -!- moltar [~me@anon.iip] has left #iip-dev [moltar]
18:17 &lt;@nop&gt; well
18:17 &lt;@nop&gt; not exactly
18:17 &lt;@hezekiah&gt; Uh oh.
18:17 &lt;@nop&gt; but imagine how that works
18:17 &lt;@nop&gt; and you can understand arbitrary sockets
18:17 &lt;@nop&gt; isproxy just routes
18:17 &lt;@nop&gt; and delivers
18:18 &lt;@nop&gt; now jrand0m
18:18 &lt;@nop&gt; quick question
18:18 &lt; jrand0m&gt; si sr?
18:18 &lt;@nop&gt; is this api designed for only new applications that are designed
to work on this network
18:18 -!- mode/#iip-dev [+v logger] by hezekiah
18:18 &lt; WinBear&gt; nop: with the highlevel replacing the irc client?
18:18 &lt; jrand0m&gt; nop&gt; yes. though a SOCKS5 proxy could use this API as well
18:18 &lt;@nop&gt; or can it be able to have a middle man that can allow already
standard clients
18:18 &lt;@nop&gt; for instance
18:19 &lt;@nop&gt; so all we would have to do is write the middleman -&gt; api
18:19 &lt; jrand0m&gt; (but note that there's no 'lookup' service available -
no DNS for this network)
18:19 &lt; jrand0m&gt; correct
18:19 &lt;@nop&gt; so that we can support say Mozilla etc
18:19 &lt;@nop&gt; so they can just code plugins
18:19 &lt; jrand0m&gt; nop&gt; yes
18:19 &lt;@nop&gt; ok
18:19 &lt;@nop&gt; or transports :)
18:20 &lt; jrand0m&gt; (e.g. the SOCKS5 has the HTTP outproxies hardcoded to
destination1, destination2, and destination3)
18:20 &lt;@nop&gt; ok
18:20 &lt; WinBear&gt; i think i get it
18:21 &lt; jrand0m&gt; w00t
18:21 &lt; jrand0m&gt; ok, one of the things I had to think about in this design
was keeping the private keys in the app's memory space - the router never
gets a hold of destination private keys.
18:21 &lt;@hezekiah&gt; So the application can send raw data over the I2P network
by sending it to the API, and it doesn't need to worry about the rest.
18:22 &lt;@hezekiah&gt; Right?
18:22 &lt; jrand0m&gt; that means the APIs need to implement the end to end part
of the crypto
18:22 &lt; jrand0m&gt; exactly hezekiah
18:22 &lt;@hezekiah&gt; OK.
18:22 &lt;@nop&gt; yes
18:22 &lt;@nop&gt; that's the idea
18:22 &lt;@nop&gt; it does it for you
18:22 &lt;@nop&gt; you just call the hook
18:23 &lt;@hezekiah&gt; One quick question:
18:23 &lt;@hezekiah&gt; This 'router' obviously needs to speak a certain protocol
over it's transports.
18:23 &lt; jrand0m&gt; correct
18:23 &lt;@hezekiah&gt; So it is possible to provide multiple implementations of
the router ...
18:23 &lt; jrand0m&gt; yes
18:24 &lt;@hezekiah&gt; ... as long as they both speak the same protocol.
18:24 &lt; jrand0m&gt; (which is why the spec has placeholders for bitbuckets)
18:24 &lt; jrand0m&gt; right
18:24 &lt;@hezekiah&gt; So you have a router in Java, and one in C, and one
in PASCAL.
18:24 * jrand0m cringes
18:24 &lt; jrand0m&gt; but yeah
18:24 &lt;@hezekiah&gt; And they all can talk together since they're talking over
TCP/IP using the same protocol.
18:24 * WinBear jumps
18:24 &lt;@hezekiah&gt; jrand0m: And yes. I don't remember my PASCAL days overly
fondly either.
18:25 &lt; jrand0m&gt; well, Pascal can talk to the C one through the TCP transport,
and the C one can talk to the Java one over the HTTP transport, for example
18:25 &lt;@hezekiah&gt; Right.
18:25 &lt; jrand0m&gt; (transports talk to other like transports, routers manage
the messages delivered between them but don't deal with how they're delivered)
18:26 &lt;@hezekiah&gt; The point I was looking to make was that the protocol is the
same, so it doesn't matter what language someone's router is implemented in.
18:26 &lt; jrand0m&gt; right
18:26 &lt;@hezekiah&gt; Cool.
18:26 &lt; jrand0m&gt; now you understand why I said "who cares" to all the C vs
Java vs etc debates? :)
18:26 &lt;@hezekiah&gt; Yup.
18:26 &lt;@hezekiah&gt; lol
18:27 &lt;@hezekiah&gt; I've got to hand it to you jrand0m. This will make it very
kind for develoeprs to write programs for this network.
18:27 &lt; jrand0m&gt; heh, well, the API ain't quite original. this is how
Message Oriented Middleware (MOM) works
18:27 &lt;@hezekiah&gt; And you could even make routers that specialize in certain
platform specific features (like 64-bit CPU's).
18:28 &lt; jrand0m&gt; absolutely
18:28 &lt;@hezekiah&gt; jrand0m: Humble too! ;-)
18:28 &lt;@hezekiah&gt; Well, it looks good to me.
18:28 &lt; jrand0m&gt; ok, UserX, nop, does this seperation make sense?
18:28 &lt;@nop&gt; of course
18:28 &lt;@nop&gt; is userx still here
18:29 &lt;@hezekiah&gt; He's been idle for 1:26.
18:29 &lt; jrand0m&gt; 'k. so then we have two tasks: design the network, and
design how the API works.
18:29 &lt;@nop&gt; right
18:29 &lt;@hezekiah&gt; Quick simple question: The API's do end to end crypto. Do
the routers to node to node crypto ?
18:29 &lt;@nop&gt; yes
18:30 &lt; jrand0m&gt; yes
18:30 &lt; jrand0m&gt; (transport level)
18:30 &lt;@hezekiah&gt; Good. :)
18:30 &lt;@nop&gt; hezekiah: it's very similar to what we have so far
18:30 &lt;@nop&gt; in that aspect
18:31 &lt; jrand0m&gt; ok.. er, shit, thecrypto aint around for comments on the
performance model.
18:31 &lt; Neo&gt; and for the paranoid, the apps can do the pgp encryption before
it hits the API ;)
18:31 &lt; jrand0m&gt; absolutely neo
18:31 &lt; jrand0m&gt; I was even tempted to leave the end to end crypto out of
the API and leave it up to the apps...
18:31 &lt;@hezekiah&gt; jrand0m: That would be cruel.
18:31 &lt; jrand0m&gt; heheh
18:32 &lt;@hezekiah&gt; BTW, the API's and the router communicate via sockets.
18:32 &lt;@hezekiah&gt; On UNIX will they be using UNIX sockets or local TCP/IP
sockets?
18:32 &lt; jrand0m&gt; prolly just local tcp/ip for simplicity
18:32 &lt;@nop&gt; hold
18:32 &lt;@hezekiah&gt; (I suppose you could make a router that accepts both.)
18:33 * hezekiah is really liking this interchangable parts setup
18:33 &lt;@nop&gt; if you hold on a sec
18:34 &lt;@hezekiah&gt; Holding ... :)
18:34 &lt;@nop&gt; I'll call thecrypto at his house
18:34 &lt;@nop&gt; see if he can get on
18:34 &lt; jrand0m&gt; hehe word
18:34 &lt;@hezekiah&gt; lol
18:34 * hezekiah dons a thick Itallian accent
18:34 &lt;@hezekiah&gt; Nop ha' got ... CONNECTIONS!
18:34 &lt; jeremiah&gt; lo
18:34 &lt;@nop&gt; hey jeremiah
18:35 &lt; jrand0m&gt; heya jeremiah
18:35 &lt;@nop&gt; would you be willing at the api level to assist with a python api
18:35 &lt; jeremiah&gt; sure
18:35 * jeremiah reads backlog
18:35 &lt; jrand0m&gt; heh word
18:35 * nop is calling
18:36 &lt;@nop&gt; he's not home
18:36 &lt;@nop&gt; he'll be back in an hour
18:36 &lt; jrand0m&gt; 'k, has anyone else read the .xls and/or have comments on
the model?
18:37 &lt;@hezekiah&gt; I read the .xls ... but I don't know much about p2p so
most of it was over my head.
18:37 &lt;@hezekiah&gt; UserX is good at that stuff.
18:37 &lt;@nop&gt; I have to read it still
18:37 &lt; jrand0m&gt; (btw, morphmix had some insane numbers... they were saying
they could expect random hosts on the net to have average 20-150ms ping times,
rather than the 3-500 I was expecting)
18:37 &lt; jrand0m&gt; coo'
18:37 &lt;@nop&gt; it's staroffice or openoffice?
18:37 &lt; jrand0m&gt; openoffice, but I exported it to .xls
18:37 &lt;@nop&gt; which is excell?
18:37 &lt; jrand0m&gt; correct
18:38 &lt;@hezekiah&gt; BTW, concerning the API ...
18:38 &lt; jrand0m&gt; si sr?
18:38 &lt;@hezekiah&gt; ... in C the boolean would be int.
18:38 &lt;@nop&gt; which email
18:38 &lt;@nop&gt; hezekiah: yes
18:38 &lt;@hezekiah&gt; The classes would be sent as structure pointers.
18:38 &lt;@nop&gt; unless you typedef boolean
18:39 &lt;@hezekiah&gt; And the functions that use byte[] would use a void* with
an additional parameter that specefies the length of the buffer.
18:39 &lt;@nop&gt; hezekiah: you're being picky :)
18:39 &lt; jrand0m&gt; nop&gt; I cant access the archives so I'm not sure what the
subject line was, but it was last week...
18:39 &lt;@nop&gt; save it for a later time
18:39 &lt;@hezekiah&gt; nop: Picky?
18:39 &lt; jrand0m&gt; heh, yeah, y'all working on the C api can work that detail out
18:39 * jeremiah is done reading backlog
18:39 &lt;@nop&gt; what's the file called
18:39 &lt;@hezekiah&gt; nop: I'm just trying to find all the stuff that is different,
so we can hammer it out like jrand0m asked.
18:40 &lt;@hezekiah&gt; I'm trying to be helpful. :)
18:40 &lt;@nop&gt; hezekiah: yes, probably off meeting time
18:40 &lt; jrand0m&gt; nop&gt; simple_latency.xls
18:40 &lt;@hezekiah&gt; boolean sendMessage(Destination dest, byte[] payload);
18:40 &lt;@hezekiah&gt; would be
18:40 &lt;@hezekiah&gt; int sendMessage(Destination dest, void* payload, int length);
18:40 &lt;@hezekiah&gt; .
18:40 &lt;@hezekiah&gt; byte[] recieveMessage(int msgId);
18:40 &lt;@hezekiah&gt; that could either be:
18:41 &lt;@hezekiah&gt; void* recieveMessage(int msgId, int* length);
18:41 &lt;@hezekiah&gt; or
18:41 &lt;@nop&gt; jrand0m: got it
18:41 &lt;@hezekiah&gt; void recieveMessage(int msgId, void* buf, int* length);
18:41 &lt;@hezekiah&gt; or
18:41 &lt; jrand0m&gt; hezekia: why not typedef struct { int length; void* data;
} Payload;
18:41 &lt;@hezekiah&gt; DataBlock* recieveMessage(int msgId)l
18:41 &lt;@hezekiah&gt; DataBlock* recieveMessage(int msgId);
18:41 &lt; jeremiah&gt; where's this xls?
18:41 &lt;@nop&gt; oh iip-dev
18:41 &lt;@hezekiah&gt; jrand0m: The struct you just mentioned is basically what
DataBlock is.
18:42 &lt; jrand0m&gt; word hezekiah
18:42 &lt;@nop&gt; subject more models
18:42 &lt;@hezekiah&gt; Chances are the C version would have DataBlocks.
18:43 &lt;@hezekiah&gt; Beyond that the only other thing to note is that each
'interface' would just be a set of functions.
18:43 &lt;@hezekiah&gt; nop: Did I find all the differences that would exist in
a C API?
18:43 &lt; jrand0m&gt; right. perhaps #include "i2psession.h" or something
18:43 &lt; jeremiah&gt; is there a mockup python api?
18:44 &lt; jrand0m&gt; no jeremiah, I don't really know python :/
18:44 &lt;@nop&gt; I would have to re-review the java api, but I would say that
you're right on target
18:44 &lt; jrand0m&gt; but it would probably be similar to the java, as python is OO
18:44 &lt; jeremiah&gt; cool, i can derive one from the C one
18:44 * nop is not a java head
18:44 &lt; jrand0m&gt; cool jeremiah
18:44 &lt; jeremiah&gt; is the c api in the thing you sent out a few days ago?
18:44 &lt;@hezekiah&gt; Yeah. Python should be able to handle the Java api.
18:44 &lt; jrand0m&gt; jeremiah&gt; that was the Java one
18:45 &lt; jrand0m&gt; oh, the Java one was today
18:45 &lt; jrand0m&gt; the older one was language independent
18:45 &lt;@hezekiah&gt; Hmm
18:45 &lt;@nop&gt; UserX says he should be able to assist with C api
18:45 &lt; jrand0m&gt; word
18:45 &lt;@nop&gt; he's busy at work at the moment
18:46 &lt; jrand0m&gt; coo'
18:46 &lt;@hezekiah&gt; One last note: With the C api, each function would probably
take a structure* to the structure that it is an 'interface' of in Java.
18:46 &lt;@nop&gt; hezekiah: loos good
18:46 &lt;@nop&gt; looks good
18:46 &lt;@hezekiah&gt; I2PSession createSession(String keyFileToLoadFrom,
Properties options);
18:46 &lt;@hezekiah&gt; would be:
18:46 &lt;@nop&gt; java and their non-native data types
18:46 &lt;@hezekiah&gt; I2PSession* createSession(I2PClient* client, char*
keyFileToLoadFrom, Properties* options);
18:46 &lt;@nop&gt; ;)
18:46 &lt; jrand0m&gt; hehe
18:46 &lt; jrand0m&gt; right hezekiah
18:47 &lt; jeremiah&gt; are we addressing unicode?
18:47 &lt;@hezekiah&gt; Anyway, if you can live with those differences, the C and
Java API's should be identical beyond that.
18:47 &lt;@hezekiah&gt; nop? Unicode? :)
18:47 &lt; jrand0m&gt; UTF8 if not UTF16
18:48 &lt;@hezekiah&gt; Perhaps Unicode should be dealt with on the application
level.
18:48 &lt; jrand0m&gt; right, charset is all the content of the message
18:48 &lt;@hezekiah&gt; Oh.
18:48 &lt; jeremiah&gt; ok
18:48 &lt;@hezekiah&gt; Java String's are done in Unicode, aren't they jrand0m?
18:48 &lt; jrand0m&gt; the bitbuckets'll all be bit defined
18:48 &lt; jrand0m&gt; yes hezekiah
18:48 &lt; jrand0m&gt; (unless you explicitly instruct them to change charsets)
18:49 &lt;@hezekiah&gt; So the string sent to the Java API would be different than
the one sent to the C API unless the C API implements strings using Unicode.
18:49 &lt; jrand0m&gt; not relevent
18:49 &lt;@hezekiah&gt; OK.
18:49 &lt; jrand0m&gt; (app-&gt;API != API-&gt;router. we only define API-&gt;router)
18:49 &lt;@hezekiah&gt; What I'm saying is this, jrand0m:
18:50 &lt;@hezekiah&gt; If I set my password with the Java API, it goes to the
router out someplace else.
18:50 &lt; jrand0m&gt; password? you mean you create a Destination?
18:50 &lt;@hezekiah&gt; Then it find another router, which sends it to another API
(?) which is implemented in C.
18:50 &lt;@hezekiah&gt; void setPassphrase(String old, String new);
18:50 &lt;@hezekiah&gt; That function.
18:51 &lt; jrand0m&gt; hezekiah&gt; thats the administrative password to access the
administrative methods of the router
18:51 &lt;@hezekiah&gt; Ah
18:51 &lt;@hezekiah&gt; Do any functions in the API which use Java String's end
up with that String being sent to another API?
18:51 &lt; jrand0m&gt; 99.9% of apps will only use I2PSession, not I2PAdminSession
18:51 &lt;@nop&gt; also, anything carried with the router gets converted for
network travel correct?
18:51 &lt;@hezekiah&gt; If so, we should probably use Unicode.
18:51 &lt;@nop&gt; unicode wouldn't be releavant
18:52 &lt; jrand0m&gt; hezekiah&gt; no. all inter-router info will be defined by
bit buckets
18:52 &lt;@hezekiah&gt; OK.
18:52 &lt; jrand0m&gt; correct nop, at the transport level
18:52 &lt;@hezekiah&gt; (I'm assuming a bit bucket is just a binary buffer, right?)
18:53 &lt; jrand0m&gt; a bit bucket is a statement that the first bit means X,
the second bit means Y, bits 3-42 mean Z, etc
18:53 &lt; jrand0m&gt; (e.g. we may want to use X.509 for the certificates bitbucket)
18:53 &lt;@hezekiah&gt; I've never dealt with that before.
18:54 &lt;@hezekiah&gt; I'll worry about it when I get there. :)
18:54 &lt; jrand0m&gt; heh word
18:55 &lt; jrand0m&gt; ok, the four things I wanted us to hit today: *router
architecture, *performance model, *attack analysis, *psyc. We've done
the first, thecrypto is offline so perhaps we delay this (unless you have
thoughts on the model nop?)
18:57 &lt;@hezekiah&gt; Um ... jrand0m. I have yet another question.
18:57 &lt; jeremiah&gt; jrand0m: where's the latest version of the network spec? is
it what you sent out on the 13th?
18:57 &lt; jrand0m&gt; si sr?
18:57 &lt;@hezekiah&gt; Well the router architecture has the API's handle keys
/sent to them by the Application/.
18:57 &lt; jrand0m&gt; jeremiah&gt; yes
18:57 &lt;@nop&gt; I don't at this time
18:58 &lt;@hezekiah&gt; Now ... the only way I see that the API gets the key is
from createSession.
18:58 &lt; jrand0m&gt; hezekiah&gt; the router gets public keys and signatures,
not private keys
18:58 &lt; jrand0m&gt; right
18:58 &lt;@hezekiah&gt; But that requires a file.
18:58 &lt; jrand0m&gt; the keys are stored in a file or in the API's memory
18:58 &lt; jrand0m&gt; yes
18:58 &lt;@hezekiah&gt; Now if the application generates a key, why can't it just
send it to the API via a buffer?
18:59 &lt;@hezekiah&gt; Must it really store it in a file, and then provide the
file name?
18:59 &lt; jrand0m&gt; no, it can be in memory if you'd like
18:59 &lt;@hezekiah&gt; There is not function to all that in the API though.
18:59 &lt;@hezekiah&gt; It's just a thought.
19:00 &lt;@hezekiah&gt; If the key is supposed to be generated only once and used
many, many times (like GPG keys), then a file makes sense.
19:00 -!- mihi [none@anon.iip] has quit [bye all, it's getting late...]
19:00 &lt;@hezekiah&gt; But if it will be generated more often, then perhaps some
way to directly send it to the API via a structure or buffer of some sort
might be nice
19:00 &lt;@hezekiah&gt; .
19:01 &lt; jrand0m&gt; yes, its generated once and only once (unless you're wearing
a tinfoil hat)
19:02 &lt; jrand0m&gt; though the createDestination(keyFileToSaveTo) lets you
create that key
19:02 &lt;@hezekiah&gt; OK.
19:02 &lt;@hezekiah&gt; So there's really no need for transfer directly from the
App to the API. A file will suffice.
19:03 &lt;@hezekiah&gt; So where were we before I so rudely interupted? :)
19:06 &lt; jeremiah&gt; so right now we're just working on the router API, not
the client one, right?
19:06 &lt; jrand0m&gt; well, we're skipping on performance analysis for now
(hopefully we can get some chatter re: it on the mailing list before next
week?). and probably the same wrt attack analysis (unless anyone read the
new spec and has comments)
19:07 &lt;@hezekiah&gt; So we're since we're skipping that, what are we supposed
to be talking about now?
19:07 &lt;@hezekiah&gt; Psyc?
19:07 &lt; jrand0m&gt; unless anyone else has other comments to bring up...?
19:08 &lt;@hezekiah&gt; Well, for once, my comment hole (also notoriously known
as my mouth) is empty.
19:08 &lt; jrand0m&gt; hehe
19:09 &lt; jrand0m&gt; ok, anyone have any thoughts on how the IRC side of things
will work, and whether psyc may be relevent or useful?
19:09 &lt; jeremiah&gt; sidenote (that pissed me off): wired's "Wired, Tired,
Expired" list had Waste as 'wired'
19:09 &lt; jrand0m&gt; heh
19:09 &lt; jrand0m&gt; do you realize how much we're going to blow everyone away?
19:09 &lt; jeremiah&gt; yep
19:09 &lt;@hezekiah&gt; jrand0m: That assumes we get this to work.
19:10 &lt; jrand0m&gt; I guarantee it will work.
19:10 &lt;@hezekiah&gt; There are a lot of other failed efforts out there.
19:10 &lt; jrand0m&gt; I quit my job to work on this.
19:10 &lt;@hezekiah&gt; Then we're going to blow everyone away. :)
19:10 &lt;@hezekiah&gt; Yeah. How is bread getting on the table when you do that?
19:10 &lt;@hezekiah&gt; GPL code doesn't pay well. ;-)
19:10 &lt; jrand0m&gt; heh
19:11 &lt;@hezekiah&gt; As for psyc ... let me put it this way:
19:11 &lt;@hezekiah&gt; The first time I heard of it was when you emailed us
about it.
19:11 &lt; jrand0m&gt; shit, I wasn't the one who found it :)
19:11 &lt;@hezekiah&gt; However, IRC is probably one of the most (if not /the/
most) prolific chat protocols around.
19:11 &lt;@hezekiah&gt; People will want IRC apps LONG before they even /know/
what psyc is.
19:11 &lt;@hezekiah&gt; jrand0m: Oops. Sorry. I forgot that detail. :)
19:12 &lt; jrand0m&gt; not according to psyc. their history goes back to 86 I think
19:12 &lt;@hezekiah&gt; The point is that the supperiority of the protocol, isn't
really as relevant as to who uses it.
19:12 &lt;@hezekiah&gt; Their _history_ may go back that far.
19:12 &lt;@hezekiah&gt; But how many people _use_ Psyc?
19:12 &lt; jeremiah&gt; yeah if they've been around since a year after I was born
(ahem) and they aren't that big yet
19:12 &lt;@hezekiah&gt; My point is that even if it's a better protocol, most
people _use_ IRC.
19:13 &lt;@hezekiah&gt; We can make the best I2P network on the planet ...
19:13 -!- Ehud [logger@anon.iip] has quit [Ping timeout]
19:14 &lt; jeremiah&gt; can someone explain briefly why we care? I thought IRC
would only be one possible application but that the network is flexible to
support psyc as well if it wanted to
19:14 &lt;@hezekiah&gt; Right.
19:14 &lt;@hezekiah&gt; Psyc can be made ...
19:14 &lt;@hezekiah&gt; ... but I'm saying we should do IRC first because more
people use it.
19:14 &lt;@hezekiah&gt; jrand0m, we can make a great I2P network, but people won't
use it unless it has something they want.
19:14 &lt; jrand0m&gt; jeremiah&gt; the reason psyc is interesting is that we may
want to implement IRC in the same vein that psyc works
19:15 &lt;@hezekiah&gt; Hence we should provide them with a 'killer-app'.
19:15 &lt; jeremiah&gt; ok
19:15 &lt; jrand0m&gt; right, IIP is invisible IRC project, and will allow people
to run IRC
19:16 &lt; jrand0m&gt; with no central server (or any server at all, actually),
theres a lot of thinking to be done to figure out how IRC will work.
psyc has a possible answer to that
19:16 &lt; jrand0m&gt; though there are others
19:17 &lt;@hezekiah&gt; As I said, psyc might do better, but people want to use IRC,
not psyc.
19:17 &lt; jrand0m&gt; and they will
19:17 &lt; jrand0m&gt; they'll use irc
19:17 &lt;@hezekiah&gt; It's all about marketing, baby! ;-)
19:17 &lt; jeremiah&gt; I'll try to read the spec and some stuff on psyc tonight
19:17 &lt; jrand0m&gt; word
19:17 &lt;@hezekiah&gt; lol
19:17 &lt; jeremiah&gt; planning to meet at 5:00 UTC tommorow?
19:17 &lt;@hezekiah&gt; No?
19:18 &lt; jeremiah&gt; or whenever
19:18 &lt; jrand0m&gt; I'm on iip 24x7 :)
19:18 &lt; jeremiah&gt; yeah but i eat
19:18 &lt;@hezekiah&gt; jrand0m: I noticed.
19:18 &lt; jrand0m&gt; 05:00 utc or 17:00 utc?
19:18 &lt;@hezekiah&gt; jeremiah: LOL!
19:18 &lt;@hezekiah&gt; Well the iip-dev meeting officially starts at 21:00 UTC.
19:18 -!- Ehud [~logger@anon.iip] has joined #iip-dev
19:19 &lt; jeremiah&gt; ok, i just said 05:00 UTC because I was talking out of my ass
19:19 &lt; jeremiah&gt; where's mids?
19:19 &lt;@hezekiah&gt; mids, left the project for a while.
19:19 &lt;@hezekiah&gt; Weren't you there a few meetings back?
19:19 &lt; jeremiah&gt; ok
19:19 &lt; jeremiah&gt; guess not
19:19 &lt;@hezekiah&gt; We had a goodbye party of sorts as part of the agenda.
19:19 &lt; jeremiah&gt; oh
19:20 &lt;@hezekiah&gt; OK ...
19:20 &lt;@hezekiah&gt; Is there anything still on the agenda?
19:20 * jrand0m doesn't have any left on mine
19:20 &lt; jeremiah&gt; about psyc:
19:20 &lt; jeremiah&gt; if this is a psyc feature, I know you mentioned it a
while ago
19:20 * hezekiah never had an agenda in the first placve
19:21 &lt;@hezekiah&gt; pace
19:21 &lt;@hezekiah&gt; place
19:21 &lt; jeremiah&gt; I don't think having each user send a message to every
other use in the room is s smart idea
19:21 &lt;@hezekiah&gt; There!
19:21 &lt; jrand0m&gt; jeremiah&gt; so you'd have redundant nominated pseudoservers
redistribute the messages?
19:21 &lt; jrand0m&gt; (pseudoservers = peers in the channel who have the list
of users)
19:21 &lt; jeremiah&gt; I don't think 'broadcasting' is that smart either, but it
seems like it'll require a _lot_ of bandwith for a given user who may be on
a modem, and with the lag from sending say... 20 messages separately would
screw up conversation
19:21 &lt; jeremiah&gt; I don't know the best solution, maybe that would be one
19:22 &lt; jeremiah&gt; I think direct messaging would be good if you wanted it,
but there are cases where it's probalby not that important
19:22 &lt;@hezekiah&gt; The message would need to be signed by the authors private
key to garuntee authenticity.
19:22 &lt;@hezekiah&gt; Though this issue won't matter for a long time still,
I think jeremiah has a point
19:22 &lt; jrand0m&gt; hezekiah&gt; that requires users wanting provable comm :)
19:23 &lt; jrand0m&gt; definitely.
19:23 &lt;@hezekiah&gt; If I had to send a message to 100 users in a channel ...
19:23 &lt; jeremiah&gt; although my average message is only a few hundred bytes,
so sending it to hundreds of users might not be so hard
19:23 &lt;@hezekiah&gt; ... well, my conversation would be /very/ slow.
19:23 &lt; jeremiah&gt; especially if you didn't wait for a response
19:23 &lt;@hezekiah&gt; 20K to send one message.
19:23 &lt;@hezekiah&gt; I don't think so. :)
19:23 &lt; jrand0m&gt; well, if there are 100 users in a channel, *someone* has
to send out 100 messages
19:23 &lt; jeremiah&gt; it's 20k?
19:23 &lt; jeremiah&gt; oh, right
19:23 &lt;@hezekiah&gt; 200 users
19:24 &lt; jeremiah&gt; hmm
19:24 &lt; jeremiah&gt; wouldn't the routers be good at that?
19:24 &lt; jeremiah&gt; we can somewhat safely assume they have decent bandwith,
right?
19:24 &lt;@hezekiah&gt; I thought each person had a 'router implementation'
19:24 &lt; jrand0m&gt; not really. if there are relays, the nomination mechanism
needs to take that into consideration
19:24 &lt; jrand0m&gt; yes hezekiah
19:24 &lt; jeremiah&gt; i haven't read the spec
19:25 &lt; jrand0m&gt; a router is your local router
19:25 &lt;@hezekiah&gt; Ugh!
19:25 &lt;@hezekiah&gt; I'm still mixing your nicks up!
19:25 &lt;@hezekiah&gt; lol
19:25 &lt; jrand0m&gt; hehe
19:25 &lt;@hezekiah&gt; Um ... where'd nop go?
19:25 &lt;@hezekiah&gt; Oh.
19:26 &lt;@hezekiah&gt; He's still here.
19:26 &lt;@hezekiah&gt; I thought he was gone for a moment,
19:26 &lt; jrand0m&gt; but jeremiah is right, psyc has some ideas we may want to
consider, though we may want to reject them
19:26 &lt;@hezekiah&gt; Let's just get the network running first.
19:26 * jrand0m drinks to that
19:26 &lt;@hezekiah&gt; If you strech your vision to the finish line, you'll trip
over the rock 3 inches in front of you.
19:27 * jeremiah feels inspired
19:27 &lt;@hezekiah&gt; lol
19:27 &lt; jrand0m&gt; I think what would be really great if we could aim to review
the network spec by next week, sending out emails to iip-dev whenever anyone
has thoughts or comments. am I out of my mind?
19:27 &lt;@hezekiah&gt; nop? Do you have anything else to add to the agenda,
or do we adjurn?
19:27 &lt;@hezekiah&gt; jrand0m: Well, I don't know if I could read all that by
next week, but I can try. :)
19:27 &lt; jrand0m&gt; heh
19:28 &lt; jrand0m&gt; its a grueling 15 pages ;)
19:28 &lt;@hezekiah&gt; 15 pages?
19:28 &lt;@hezekiah&gt; It looked more like 120!
19:29 &lt; jrand0m&gt; heh, well, depends on your resolution I suppose ;)
19:29 &lt; jeremiah&gt; he has a lot of anchors in there, makes it look like
it's huge
19:29 &lt; jrand0m&gt; hehe
19:29 &lt;@hezekiah&gt; The left side has a LOT more than 15 links, budy!
19:29 &lt;@hezekiah&gt; 'Fess up!
19:29 &lt;@hezekiah&gt; It's more than 15. :)
19:29 &lt;@hezekiah&gt; Oh!
19:29 &lt;@hezekiah&gt; Those aren't pages! They're just anchors!
19:29 &lt;@hezekiah&gt; I'm saved!
19:30 * hezekiah feels like a seaman just rescued from drowning
19:30 &lt; jeremiah&gt; class turn to volume 4 chapter 2 Message Byte Structure
19:30 &lt; jrand0m&gt; lol
19:30 &lt;@hezekiah&gt; lol
19:30 &lt;@nop&gt; adjourn
19:30 &lt;@hezekiah&gt; *baf*!
19:30 &lt;@hezekiah&gt; Next week, 21:00 UTC, same place.
19:30 &lt;@hezekiah&gt; See y'all there. :)
19:30 &lt; jeremiah&gt; seeya
--- Log closed Tue Jul 15 19:30:51 2003
</pre>