I2P dev meeting, October 11, 2005

Quick recap

  • Present:

bar, cervantes, Complication, dust, jrandom, Myo9, postman, redzara, wiht,

IRC 完整日志

16:29 < jrandom> 0) hi
16:29 < jrandom> 1)
16:29 < jrandom> 2) I2PTunnelIRCClient
16:29 < jrandom> 3) Syndie
16:29 < jrandom> 4) I2Phex
16:29 < jrandom> 5) Stego and darknets (re: flamewar)
16:29 < jrandom> 5) ???
16:29 < jrandom> 0) hi
16:29 <@cervantes> (6)
16:29 <+postman> you mean 6)?
16:29 < jrandom> yeah, i can't count ;)
16:30  * postman high-fives cervantes 
16:30 < jrandom> weekly status notes posted up @ http://dev.i2p.net/pipermail/i2p/2005-October/000990.html
16:30 < wiht> Questions should be item 6.
16:30 < jrandom> since i'm 30 minutes late, y'all have already read through those notes, I'm sure, so lets get this underway ;)
16:31 < jrandom> 1)
16:31 <@cervantes> 6) Discuss jrandom's roommate's poor judgement in timing
16:31 < jrandom> *cough* ;)
16:31 < jrandom> ok, as mentioned in the email, the release seems to be doing pretty well
16:32 < jrandom> we've found the bug that had kept the irc servers back on an older build, and they're now up to date too (w00t!)
16:32 <+postman> :)
16:32 < wiht> Speaking of that, in the netDB on the router console, would it be possible to list the table with routers and their versions at the top of the page?
16:33 < jrandom> the number of routers per version, right?  sure, that could be done pretty easy, maybe integrate it into the peers.jsp table (showing per-peer version) and a new table at the bottom?
16:34 < jrandom> its kind of nice seeing 9 versions playing well together, though of course newer ones work best
16:35 < jrandom> ok, anyone else have something to bring up regarding 1)
16:35 <+postman> one of my routers shows 1080 known
16:35 < jrandom> zounds
16:35 <+postman> i think this is a bit off track?
16:35 < jrandom> is that on
16:35 <+postman> yeah, think so
16:36 < jrandom> hmm, yeah, thats a... bit high.  i'm seeing about half that many right now
16:36 <+Complication> Sustainably 400-ish here
16:37 <+bar> same here
16:37 < wiht> I see 260 known routers.
16:37 < jrandom> postman: perhaps we can dig into whats going on in that router after the meeting (could you tar.bz2 me netDb/routerInfo-*?)
16:38 <+postman> jrandom: yeah thanks
16:38 < jrandom> gracias
16:38 < jrandom> yeah, not everyone will see every netDb reference, so thats normal for there to be fluctuation
16:40 < jrandom> ok, if there's nothing else on 1), lets swing on over to 2) I2PTunnelIRCClient
16:40 <@cervantes> nice one dust
16:40 < jrandom> as mentioned in the email, we've got a new IRC-protocol-specific filter available in CVS, and it should be rolled out as the default in the next rev
16:41 <+postman> cool
16:41 < jrandom> yeah, this is very cool, people have been asking for something like it for ages
16:41 <+Myo9> Jrandom, you have become more open recently, we have learned about your ex, and now your room mate, etc. Do recall: http://www.navysecurity.navy.mil/st031204.jpg
16:41 < jrandom> *cough*
16:42 < dust> if you wish to see what your client send you can add net.i2p.i2ptunnel.I2PTunnelIRCClient=INFO and then look at the logs to see it all
16:43 < dust> i've tested some clients but there are many..
16:43 < jrandom> yeah, i've been watching it for a lil bit, but the filtering seems sound
16:44 < jrandom> there are some neat things we may be able to do in the future too - e.g. PING/PONG locally, to cut down on network activity
16:44 <+Complication> dust: thanks for the "info" :)
16:44 <+bar> awsome dust, thanks a lot
16:44 < wiht> Does this mean we don't need to set up an extra IRC tunnel?
16:44 < jrandom> wiht: no, you'll need an irc tunnel, but it can replace the one you use already
16:45 <+Complication> wiht: just worry less about our IRC client giving away our ass
16:45 < jrandom> postman/cervantes: any thoughts on increasing or removing the server ping/pong timeouts?  
16:45 < wiht> That explains it, thanks.
16:46 <+postman> mmh, i would not remove them, my client completely freaked when i played around with it
16:46 < jrandom> postman: well, i'm thinking if it responded to them locally, so the client would get a really, really fast PING/PONG
16:46 <@cervantes> postman: the proxy could respond to pings
16:46 < jrandom> (but the ping/pong wouldn't need to go over the network)
16:47 < jrandom> i don't know the impact, but it may be worth looking into.
16:47 <@cervantes> but I'm not sure how the servers would react, you might end up with a bunch of zombie clients
16:47 <+postman> jrandom: well
16:47 < jrandom> well, the streaming lib's keepalive should handle that
16:47  * Complication has occasionally experienced zombification
16:47 < jrandom> Complication: recently?
16:47 <+postman> jrandom: if the proxy pings for the client, the proxy must ping/pong to the client as well
16:48 <+Complication> A week ago, I think.
16:48 < jrandom> postman: a PING from the client to the proxy would have the proxy respond directly to the client with a PONG without sending anything over i2p
16:48 <+Complication> But my "copy" was dropped eventually.
16:48 <@cervantes> jrandom: the connection would be held open...the servers would need to lower their threshold for deciding at what point a client is stale and need ejecting
16:48 < jrandom> Complication: ah, the irc servers weren't up to date back then, shouldn't happen anymore
16:49 <+Complication> Without me using "ghost". Recent uses of the ghost command have been due to operating with many nodes.
16:49 <+postman> jrandom: and the lag measurement?
16:49 < jrandom> cervantes: right.  and/or if necessary, the proxy could inject an extra PING message to the server if it /needs/ one. 
16:49 <+postman> i find it quite useful to know if i am lagged or not
16:49 < jrandom> postman: i do too, but you can always /msg yourself
16:50 < dust> you could perhaps reduce the number of pings
16:50 < jrandom> it would save a substantial amount of bandwidth, since tunnel messages are 1024byte blocks, sent over 2*k+1 hops
16:50 < jrandom> that too
16:50 < jrandom> i don't know, just an idea.  what we have now is kick ass regardless 
16:51 <+postman> ok, i would try to patch a testserver
16:51 <@cervantes> perhaps we could look into reducing the number...but I think we should still send some real pings do determine if the clients are still alive
16:51 <+postman> maybe it works
16:51 < jrandom> seems reasonable cervantes.  i don't think it'd need any patching on the server, i hope?
16:52 <+postman> jrandom: to deactivate maybe - but lowering the interval is conf parameter
16:53  * postman chews on the ircd documentation ( again )
16:53 < jrandom> cool, no rush.  just something we can look into sometime
16:53 <@cervantes> class servers
16:53 <@cervantes> {
16:53 <@cervantes>     pingfreq   120;
16:54 <@cervantes> class clients { pingfreq 90 }
16:54 <@cervantes> that's my current config
16:54 <+postman> cervantes: yes, i know  - the question is if it can be deactivated at all
16:54 <@cervantes> I wouldn't deactivate them...just look into reducing them
16:55 <+postman> ok, let's start with that
16:55 <+postman> cervantes: how about 180 secs?
16:56 <@cervantes> in at the deep end with 240
16:56 <@cervantes> but perhaps we should ge tthe ircproxy side of things ready first
16:57 <@cervantes> *discuss after meeting*
16:57 <+postman> agreed
16:57 < jrandom> w3rd.  ok, anything else on 2) I2PTunnelIRCClient, or shall we move on to 3) Syndie?
16:57 <@cervantes> anything to reduce my current 40kb/sec average router traffic ;-)
16:58 < jrandom> heh, for some reason i doubt thats all irc ;)
16:58 < jrandom> ok, movin' on
16:59  * cervantes hides is pony video downloads he's been leeching from jrandom all week
16:59 <@cervantes> is=the
16:59 <+postman> LOL
16:59 < jrandom> as mentioned in the mail, there's some pretty cool stuff going on with syndie
16:59 < jrandom> the cli is trivial, but dust's new Sucker looks really promising
16:59 < jrandom> dust: wanna give us a rundown?
17:00 < dust> oh,
17:01 < dust> well, it uses rome for parsing the feeds and then converts it to sml, like described in jrandoms blog
17:02 < dust> it's not what you'd call robust yet, but it's only two days old :)
17:02 < dust> i've got some dilbert in my syndie..
17:02 < dust> :)
17:02 < dust> .
17:02 < jrandom> nice
17:03 < jrandom> ok, what are your thoughts on where its going - should we toss it into the syndie source and expose it as a cli, or keep it separate and distribute it independently, or something else?
17:04  * dust don't know, you decide
17:04 < dust> the less separate tools the better
17:04 < jrandom> yeah, probably easier to bundle it all together, that way everyone knows they can use it
17:05 < jrandom> we'd then be able to do things like integrate it into the web interface, and maybe into Ragnarok's scheduler (syndicating with other nodes and pulling from rss/atom/etc)
17:07 < jrandom> ok, anyone have any questions/comments/concerns on 3) Syndie?
17:07 < wiht> If you keep integrating software into I2P, it may become a bloated software package.
17:07 < wiht> Of course, I can turn off Syndie if I am not using it.
17:08 < jrandom> the i2p sdk 13KLOC
17:08 < jrandom> and the i2p router is only 22KLOC
17:08 < jrandom> but yeah, there is an impact on download times of the install
17:09 < jrandom> if someone wanted, they could build a stripped down router with no client apps, using just the router.jar, jbigi.jar, and i2p.jar
17:09 < wiht> Yes, I was referring to the download.
17:09 < jrandom> (but its much more useful when there's a web interface to control it, and i2ptunnel, and the streaming lib, etc ;)
17:11 < jrandom> smeghead was working on a distribution system (like emerge, for java), and there's the jpackage folks too
17:11 < jrandom> if someone wants to look into a seamless and reliable way to manage the apps without bundling, it'd be pretty cool
17:12 < jrandom> ok, if there's nothing else on that, lets jump on over to 4) I2Phex
17:13 < jrandom> I don't really have much to add beyond whats in the status notes
17:13 < jrandom> redzara: you around?
17:13 <+redzara> yes i'm
17:13 <+redzara> I'm already working on the next version, while waiting for the meeting with Gregor.
17:13 < jrandom> ah great
17:13 <+redzara> Work, for the moment, primarily consists in identifying the differences and the needs related to the use of I2P such as for example tcp/udp vs i2p, management of the parameters specific to I2P (and management of the update of these same parameters at the time of the next versions, ...) port of GWebCache to I2P, use RSS or not, use push or not... 
17:14 <+redzara> I have much documentation and code to read
17:15 < jrandom> wow, yeah, sounds like a lot.  let me know if you've got any questions regarding i2p integration, or if you just want someone to bounce ideas off
17:16 < jrandom> getting the I2Phex part into a plugin for the mainline Phex would be really kickass
17:17 < jrandom> ok, anyone else have anything for 4) I2Phex?
17:18 <+redzara> I would need certainly assistance for the petname part
17:19 <+redzara> and maybe too for fine tunning tunnels's parameters
17:19 < jrandom> cool, the naming is pretty easy - at a basic level, you could even get by without using names at all (this is how I2Phex does it now)
17:20 < jrandom> tunnel config shouldn't be a problem either, though that brings up the idea that maybe Phex would need an "advanced configuration" section for plugins
17:20 < jrandom> (we'd obviously want to have good defaults anyway)
17:21 <+redzara> maybe something like ircclient, an filter to be sure
17:22 <@cervantes> better to get the app in shape imho
17:22 < jrandom> that might work, though dealing with arbitrary byte sequences may be tough
17:23 < jrandom> though, a proxy like ircclient might be able to allow any gnutella client to use it.  but it'd be a bunch of work.
17:23 <+redzara> humm, it's just an idea ;)
17:23  * jrandom doesn't know the protocol well enough to say what the best approach is, so suggest going with the simplest thing that could possibly work :)
17:25 < jrandom> ok, if there's isn't anything else, perhaps we can swing through 5) stego and darknets briefly
17:26 < jrandom> i'm not sure if there's anything i have to add beyond whats being said on the list (and major discussion should probably continue there)
17:27 < jrandom> that said, is there anything people want to bring up about the issues raised?
17:27 < wiht> Freenet version 0.5 and 0.7 were mentioned in the discussion. Is there a version 0.6 for Freenet?
17:27 < jrandom> 0.6 is their current "unstable" branch of the network
17:27 < jrandom> afaik
17:27 <+postman> ohh and i thought it has been stolen by alien forces
17:28 < jrandom> while blaming the aliens is usually a safe bet, this is one of the few instances where they're not at fault
17:28 <+postman> :)
17:28 < wiht> Toad was talking about being able to harvest the IP addresses of I2P or FreeNet nodes, right?
17:28 < jrandom> among other things
17:29 < wiht> Just wanted to clarify that, thanks.
17:29 < jrandom> np.  ok, anyone else have anything on 5), or shall we move on over to the good ol' fashioned 6) ???
17:30 <+postman> ok, i got one for 6)
17:30 < jrandom> consider us moved.
17:30 < jrandom> whats up postman?
17:30 <+postman> we all have seen that protocol specific filter capable proxies are good and needed
17:31 <+postman> would it be feasable to invest thinking in a generic proxy
17:31 <+postman> that can be fed with a protocol description
17:31 <+redzara> I would like to have an application like cron using beanshell to run code java code dynamically
17:31 <+postman> along with stuff to watch for/filter/disguise
17:31 <+postman> like a filter/sanitize xml description 
17:32 <+postman> so that we dont need new source but just a new filter file/profile
17:32 <+postman> (just a question if its worth to think about it)
17:32 < jrandom> very, very complicated postman.  it'd be possible to use a lexer like javacc to build input languages and an app to translate that language into the output format
17:32 <@cervantes> it's catching the stuff that deviates from the protocol that's tricky
17:33 <+postman> it was just an idea to trigger a process of brainstorming
17:33 <+postman> imho something like a generic proxy with modeled filter/parser is very usable
17:33 < wiht> Has anyone been able to connect to eepsites.i2p? I have tried several times over the past week, but was always unsuccessful.
17:33 < jrandom> wiht: i loaded it once, its the same as eepsites.com
17:34 < jrandom> (or is it .net?  or .org?  i forget)
17:34  * wiht visits eepsites.com
17:34 < jrandom> postman: if someone could come up with something that'd work, that'd kick ass
17:34 <+postman> jrandom: ok, i'll do some thinking together with susi
17:34 < jrandom> w3wt
17:34 <+postman> jrandom: maybe we'll drop it next week 
17:35 < wiht> It is eepsites.com, and it is a search engine for eepsites.
17:35 <+postman> but i had a dream that it worked
17:35 <+postman> :]
17:35 < jrandom> :)
17:36  * Complication suspects that descibing all the subtleties which occur in protocols... requires code, and nothing less than code
17:36 <+Complication> (for most protocols, at least)
17:36 <@cervantes> nah just some eeevul regex
17:36 <+postman> Complication: maybe this suspicion is the reason that keeps us from further investigation
17:37 <+postman> Complication: i am not yet sure, but suspicion alone will not put me to rest on that matter
17:37 < jrandom> well, an important point here is something dust demonstrated for us -
17:37  * Complication fears a regex capable of such things
17:37 < jrandom> code isn't necessarily that scary.
17:37 <+postman> see? :)
17:37 <+postman> a good filter modelling language will do the same
17:38 <+postman> :)
17:38 <@cervantes> tcl? :)
17:38 <+Complication> It would have to be good.
17:38  * jrandom sees that you've got your own flying ponies too postman ;)
17:38  * dust also felt bad about duplicating code here and there
17:38 <+postman> jrandom: no cows :)
17:38 < jrandom> working code >>> theoretical improvements in code 
17:39 <+postman> mmh
17:40 <+postman> one thing i learned from i2p
17:40 < wiht> >>> means "much, much better?"
17:40 <+postman> do not give up on first looks
17:40 < jrandom> true enough postman 
17:40 < jrandom> yes wiht 
17:41 < jrandom> it would be really cool
17:41 < jrandom> ok, anyone else have something to bring up for the meeting?
17:41 <+bar> well, how's the IMAP working, postman? (i read about it in the forums but haven't tried it yet myself)
17:41 <+postman> bar: try it yourself - i have no user reports yet
17:41  * cervantes rolls in the pony shaped gong
17:42 <+bar> ok, will do :)
17:42 <+postman> bar: and for me it works FINE :)
17:42 < jrandom> nice
17:42 <+bar> cool
17:42 <+postman> cervantes: you're fixated
17:42 <@cervantes> me?!
17:42 <@cervantes> :)
17:43 < jrandom> ok, before we reach the 90 minute mark
17:43  * jrandom winds up
17:43  * jrandom *baf*s the meeting closed