[00:00:14] *** neitcho has quit IRC
[00:06:40] *** BentMyWookie has joined #bittorrent
[00:07:04] *** BentMyWookie has joined #bittorrent
[00:09:15] *** Andrius has quit IRC
[00:10:59] *** BentMyWookie has quit IRC
[00:11:15] *** BentMyWookie has joined #bittorrent
[00:31:45] *** ajaya has quit IRC
[01:00:05] *** swinokur has joined #bittorrent
[01:36:59] *** TheSHAD0W has quit IRC
[01:46:59] <erdgeist> hydri: ping
[01:56:37] <hydri> erdgeist: hey
[01:57:00] *** BentMyWookie has quit IRC
[01:57:24] *** BentMyWookie has joined #bittorrent
[01:58:19] *** BentMyWookie has quit IRC
[01:58:45] *** BentMyWookie has joined #bittorrent
[01:58:52] *** BentMyWookie has quit IRC
[01:59:06] *** BentMyWookie has joined #bittorrent
[01:59:39] <erdgeist> hydri: Oh, I'm justing writing up my thoughts about multi-announces in a mail
[01:59:52] <hydri> ok
[02:00:21] *** BentMyWookie has quit IRC
[02:00:37] *** BentMyWookie has joined #bittorrent
[02:02:02] *** andar2 has quit IRC
[02:05:07] *** TheSHAD0W has joined #bittorrent
[02:14:03] <alus> erdgeist: why multi-announce and not keep-alive?
[02:15:04] <erdgeist> alus: udp knows no keep alive ;)
[02:15:08] <alus> ah
[02:15:21] <alus> well, why multi-announce with udp then?
[02:15:47] <The_8472> multi announce seems... misguided imo. people shouldn't run more than a handful of torrents at a time. clients scheduling them not all at once either (except startup)
[02:16:04] <alus> multi-scrape is much more likely, I agree
[02:16:14] <erdgeist> alus: well, what I'd love to see is announce?event=stopped for a quitting app in ONE packet
[02:16:20] <The_8472> is there a real problem that actually needs to be fixed?
[02:16:27] <erdgeist> alus: transmission takes ages quitting
[02:16:43] <alus> but is that multi or just new udp tracker semantics?
[02:16:54] <erdgeist> alus: thats multi-announce
[02:16:56] <The_8472> it's UDP ... get a transaction ID and churn out packets in a fire-and-forget manner...
[02:17:17] <The_8472> or use HTTP pipelining
[02:17:17] <alus> surely it is not the time of sending num_torrents number of udp packets vs. one
[02:17:26] <alus> it's got to be timeouts on the responses, no?
[02:17:30] <erdgeist> The_8472: since it is an announce, clients do expect an answer
[02:17:48] <alus> well that sounds like new non-response semantics, not multi-announce
[02:17:49] <erdgeist> The_8472: they even complained that opentracker sends empty answer dictionaries if event=stopped
[02:17:50] <The_8472> they can choose not to
[02:18:06] <The_8472> oh, are we talking about HTTP or UDP now?
[02:18:14] <alus> or what The_8472 said. just don't wait.
[02:18:22] <erdgeist> they complained in http
[02:18:31] <The_8472> pipelineing...
[02:18:38] <erdgeist> they won't just fire and forget unless it is spec'd
[02:18:40] <alus> but if you want to be sure the tracker got it, you have to wait. I don't see how reducing the number of packets helps very much
[02:18:48] <erdgeist> for pipe lining you need kee alive
[02:19:02] * alus is talking to thin air
[02:19:02] <erdgeist> I do see this very much
[02:19:08] <alus> see what?
[02:19:13] <erdgeist> in an environment where lots of packets get lost
[02:19:23] *** bittwist has joined #BitTorrent
[02:19:29] <alus> reducing the number of packets does not decrease the packet loss rate
[02:19:39] <erdgeist> but the chance that all info_hashes get through
[02:19:48] <alus> no
[02:19:48] <erdgeist> ... in a single packet
[02:19:50] <erdgeist> yes
[02:19:54] <The_8472> pipelining is on a different abstraction level though, it doesn't need any changes to the protocols. at most it needs guidelines when to use it and when to close connections immediately
[02:20:15] <The_8472> that seems to be a solution with the highest for/backward compatibility
[02:20:16] <erdgeist> if you use an http lib, maybe. I need to implement it by hand
[02:20:37] <erdgeist> ... already did so
[02:20:56] <alus> erdgeist: imagine all infohashes were sent in only one packet, and the other 9 packets were empty. how can you be sure you got the one with all the infohashes?
[02:20:56] <erdgeist> but no client atm supports keepalive/pipe lining
[02:21:32] <alus> multi-announce does not solve packet loss
[02:21:38] <The_8472> yeah, we close connections immediately atm because trackers didn't support it. and because there are very few situations where it's of any use
[02:21:38] <erdgeist> alus: you get a reply? also not sending 8 packets reduces the chance of it not getting lost
[02:21:58] <alus> erdgeist: not significantly, because these packets are small
[02:22:24] <alus> erdgeist: likely you are competing in a buffer with 1500 byte packets. probably either 1500 bytes are free or 0 are free
[02:22:40] <The_8472> btw, what's causing the packet loss?
[02:22:42] <alus> erdgeist: so if all 10 packets fit in the buffer, none are dropped
[02:22:51] <erdgeist> Lots of traffic causes it ;)
[02:23:00] <alus> then implement uTP
[02:23:21] <erdgeist> you don't send them at the same second, they arrive at different times, may over different routes
[02:23:29] <The_8472> well, unless links are congested traffic should not increase the loss _rate_, only the absolute amount of lost packets, by the virtue of having more traffic ^^
[02:23:33] <alus> but where are they being dropped?
[02:23:40] <alus> if it's due to the peer uploading, the bottleneck is local
[02:23:55] <erdgeist> if there's a 10% drop rate, you end up with a higher chance of a dropped packet by sending 10 packets than by sending one
[02:24:11] <alus> erdgeist: of -a- dropped packet. not of -all- dropped packets
[02:24:27] <erdgeist> alus: still you have to wait for timeout and resend the missing one
[02:24:36] <erdgeist> alus: et voila
[02:24:49] <The_8472> yes, but you don't have to do that over and over
[02:24:52] <The_8472> you can do things in parallel
[02:24:57] <alus> erdgeist: which is the same as one packet being dropped 10% of the time and having to resend the multi-announce
[02:25:00] <The_8472> so the timeouts shouldn't be additive
[02:25:32] <erdgeist> alus: you get a 30% chance of all packets going through instead of a 10% chance of ONE packet going through
[02:25:40] <The_8472> <alus> erdgeist: which is the same as one packet being dropped 10% of the time and having to resend the multi-announce <- and having to resend the multi announce 10% of the time
[02:25:52] <alus> erdgeist: what?
[02:26:01] <erdgeist> aehm, I mean 90% chance
[02:26:21] <alus> 10% drop rate means 90% chance of all packets going through
[02:26:25] <erdgeist> alus: a 10% drop rate: 90% success chance for 1 packet, 31,3% success rate for 10 packets
[02:26:32] <alus> what?
[02:27:07] <The_8472> that would be assuming you can only send 10 packets or 0
[02:27:13] <erdgeist> alus: 0.9^10 is the chance for all packets going through
[02:27:21] <The_8472> but since you can resend packets individually that calculation doesn't apply
[02:27:31] <erdgeist> but you have to WAIT for the timeout
[02:27:41] <alus> you don't care about the chance of all 10 making it through at once
[02:27:41] <The_8472> yes, but it doesn't matter how many packets you send
[02:27:43] <erdgeist> for every packet
[02:27:52] <erdgeist> alus: you do.
[02:27:53] <alus> you care about eventually getting all 10 through
[02:28:04] <alus> so you don't have to wait for the network to not drop all 10 all at once
[02:28:23] <alus> so that calculation doesn't apply
[02:28:23] <erdgeist> alus: if even only one does not get through you have to resend it
[02:28:38] <alus> erdgeist: yes, and that happens the same % of the time as the multi-announce not making it through
[02:28:46] <erdgeist> alus: if you get 1 reply for only one packet with 90% chance, all's fine
[02:29:11] <erdgeist> alus: only in 10% you have to wait to resend
[02:29:35] <alus> erdgeist: these cases are the same.
[02:29:41] <erdgeist> compared to 70% with 10 packets
[02:29:52] <alus> no, you're doing the statistics wrong
[02:29:56] <erdgeist> where?
[02:30:29] <erdgeist> If the app quits, I want it to do that as fast as possible. right?
[02:30:53] <erdgeist> So I want to reduce the chance of my app having to resend to the least amount possible
[02:30:54] <alus> 10% drop rate is 10% drop rate, no matter how many packets you send. so say you send 1 packet, it either is sent 90% of the time, or not. if you send 10, probably 9 make it through and one does not.
[02:31:01] <alus> so if 1 did not make it through you have to resend it
[02:31:08] <alus> regardless of how many you send before that
[02:31:47] <alus> so if you tried to send 1 and hit the bad 10% case, you have to resend. if you sent 10 and you hit the bad case and one didn't make it, you hanve to resend.
[02:31:55] <alus> both resends. both 1% of the time.
[02:32:04] <alus> er 10%
[02:32:11] <alus> erdgeist: see?
[02:35:24] *** swinokur has quit IRC
[02:35:24] *** mxs has quit IRC
[02:35:24] *** r3wj has quit IRC
[02:35:25] *** mpl has quit IRC
[02:35:25] *** A9[idle] has quit IRC
[02:35:25] *** init0 has quit IRC
[02:35:26] *** Koper has quit IRC
[02:35:26] *** Firon has quit IRC
[02:35:26] *** hlindhe_ has quit IRC
[02:35:26] *** wojci has quit IRC
[02:35:26] *** deltab has quit IRC
[02:35:26] *** The_8472 has quit IRC
[02:35:27] *** bittwist has quit IRC
[02:35:27] *** kwinz2 has quit IRC
[02:35:28] *** kjetilho has quit IRC
[02:35:28] *** andar has quit IRC
[02:35:28] *** MassaRoddel has quit IRC
[02:35:28] *** foomor has quit IRC
[02:35:29] *** nks has quit IRC
[02:35:30] *** void^ has quit IRC
[02:35:30] *** K`Tetch has quit IRC
[02:35:30] *** goussx has quit IRC
[02:35:31] *** burris has quit IRC
[02:35:31] *** _Mazon has quit IRC
[02:35:31] *** uriel has quit IRC
[02:35:31] *** TheSHAD0W has quit IRC
[02:35:31] *** bt42 has quit IRC
[02:35:32] *** HandheldPenguin` has quit IRC
[02:35:32] *** ProperNoun has quit IRC
[02:35:32] *** BentMyWookie has quit IRC
[02:35:33] *** alienvenom has quit IRC
[02:35:33] *** cgreco has quit IRC
[02:35:33] *** Astro has quit IRC
[02:35:34] *** Nolar has quit IRC
[02:35:34] *** GTHK has quit IRC
[02:35:34] *** jnpplf has quit IRC
[02:35:35] *** erdgeist has quit IRC
[02:35:35] *** BitTorrentBot has quit IRC
[02:35:35] *** DreadWingKnight has quit IRC
[02:35:35] *** alus has quit IRC
[02:35:36] *** charles has quit IRC
[02:35:36] *** tris has quit IRC
[02:35:37] *** swolchok has quit IRC
[02:35:37] *** DeHackEd has quit IRC
[02:35:37] *** hlindhe has quit IRC
[02:35:38] *** Gottaname has quit IRC
[02:35:39] *** bittwist has joined #bittorrent
[02:35:39] *** TheSHAD0W has joined #bittorrent
[02:35:39] *** BentMyWookie has joined #bittorrent
[02:35:39] *** swinokur has joined #bittorrent
[02:35:39] *** r3wj has joined #bittorrent
[02:35:39] *** K`Tetch has joined #bittorrent
[02:35:39] *** bt42 has joined #bittorrent
[02:35:39] *** goussx has joined #bittorrent
[02:35:39] *** Nolar has joined #bittorrent
[02:35:39] *** GTHK has joined #bittorrent
[02:35:39] *** kwinz2 has joined #bittorrent
[02:35:39] *** swolchok has joined #bittorrent
[02:35:39] *** DeHackEd has joined #bittorrent
[02:35:39] *** hlindhe has joined #bittorrent
[02:35:39] *** BitTorrentBot has joined #bittorrent
[02:35:39] *** charles has joined #bittorrent
[02:35:39] *** alus has joined #bittorrent
[02:35:39] *** tris has joined #bittorrent
[02:35:39] *** cgreco has joined #bittorrent
[02:35:39] *** Gottaname has joined #bittorrent
[02:35:39] *** jnpplf has joined #bittorrent
[02:35:39] *** Astro has joined #bittorrent
[02:35:39] *** erdgeist has joined #bittorrent
[02:35:39] *** alienvenom has joined #bittorrent
[02:35:39] *** DreadWingKnight has joined #bittorrent
[02:35:39] *** mxs has joined #bittorrent
[02:35:39] *** hlindhe_ has joined #bittorrent
[02:35:39] *** The_8472 has joined #bittorrent
[02:35:39] *** init0 has joined #bittorrent
[02:35:39] *** Firon has joined #bittorrent
[02:35:39] *** Koper has joined #bittorrent
[02:35:39] *** mpl has joined #bittorrent
[02:35:39] *** wojci has joined #bittorrent
[02:35:39] *** deltab has joined #bittorrent
[02:35:39] *** A9[idle] has joined #bittorrent
[02:35:39] *** _Mazon has joined #bittorrent
[02:35:39] *** uriel has joined #bittorrent
[02:35:39] *** HandheldPenguin` has joined #bittorrent
[02:35:39] *** burris has joined #bittorrent
[02:35:39] *** ProperNoun has joined #bittorrent
[02:35:39] *** kjetilho has joined #bittorrent
[02:35:39] *** foomor has joined #bittorrent
[02:35:39] *** nks has joined #bittorrent
[02:35:39] *** andar has joined #bittorrent
[02:35:40] *** MassaRoddel has joined #bittorrent
[02:35:40] *** void^ has joined #bittorrent
[02:35:57] <erdgeist> actually you increase your chance of losing to 100% for all the other tickets, unless they are all the same :
[02:36:00] <erdgeist> :)
[02:36:23] <alus> erdgeist: %*#@(%#^!@
[02:36:45] <The_8472> depending on your acknowledgement mechanism sending more packets actually makes things faster. if i send 1 packet at a time in my DHT implementation it's slow like molasses. doing 10 packets at once and waiting on timeouts concurrently on the other hand speeds things up.
[02:37:05] <alus> erdgeist: that is not how statistics work. you do not change the chances of one ticket by purchasing a different one
[02:37:14] *** bt42 has quit IRC
[02:37:44] <The_8472> i'd like to see a description of the _actual_ problem
[02:38:08] <alus> erdgeist: you are conflating the individual chance of one ticket with the combined result of all tickets you hold. just like the UDP packet case, you do not need to win on all of your lottery tickets to win the lottery
[02:38:15] <erdgeist> agreed :) because you're looking at winning the lottery. I am looking at independent events
[02:38:22] <TheSHAD0W> http://news.slashdot.org/article.pl?sid=10/01/19/2253213
[02:38:51] <erdgeist> I think the problem goes away, once transmission supports udp
[02:38:55] <erdgeist> and it will work out
[02:39:20] <alus> erdgeist: the UDP packet packet case is different from the lottery. it was a bad comparison to start.
[02:39:49] <alus> erdgeist: but one multi-announce is still the same as 10 packets.
[02:40:46] <hydri> alus: are you assuming that loss at a certain percentage per byte or per packet?
[02:40:57] <hydri> I believe it's typically per packet
[02:41:06] <erdgeist> so do I
[02:41:24] <hydri> in which case, sending one larger packet instead of 10 small ones, you actually gain something..
[02:41:57] * Nolar would argue "who cares if some stop announces are dropped"
[02:42:23] <erdgeist> lets see: if there is 10 soldiers in my army and 1 other soldier and the chance of getting them through a mine field is 90%
[02:42:32] <alus> hydri: well there's either space for the packet or there isn't, right?
[02:43:14] <hydri> alus: maybe.. maybe the buffer is divided into MTU sized slots, and then there's either space for a packet or no space
[02:43:30] <hydri> alus: i.e. the loss rate is dependent on the packet, not the bytes
[02:43:48] <hydri> depending on which assumptions you have you come to different conclusions..
[02:43:50] <alus> hydri: and one packet consumes one MTU slot regardless of how big it is?
[02:43:55] <hydri> right
[02:44:20] <alus> ok, it still works them
[02:44:24] <alus> er, then
[02:44:38] <The_8472> anyway, we're just speculating
[02:45:05] <The_8472> we don't even know where exactly the slowness in transmission occurs, what's causing the packet losses, etc. etc.
[02:45:34] <erdgeist> well, atm it's http connections to the tracker, some of them failing
[02:45:47] <The_8472> are they doing them parallel or one by one?
[02:47:05] <erdgeist> out of the blue: they're using libevent. So I guess, they schedule them parallel
[02:47:47] <The_8472> then most of the announces should succeed immediately, only a small amount needs to be retried
[02:48:10] <The_8472> and... the packet loss rates must be insane if TCP connections time out
[02:48:16] <erdgeist> but tcp timeout is huge
[02:48:17] <The_8472> because tcp itself already retries
[02:48:26] <The_8472> not on connection setup
[02:49:30] <The_8472> if tcp connection setups fail there's a more serious problem to fix imo
[02:49:36] <erdgeist> I guess, opening 100+ http connections in parallel uses up some system ressources
[02:49:46] <alus> hydri: the problem with expressing packet loss in terms of drop rate is that it seems to indicate sending 10 GB will drop only x% of the packets. which is not true, it will drop all after however big the buffer is
[02:50:26] <alus> hydri: so the 1 packet == 10 packets statistical model is correct if drop rate is a real thing, not an observed effect
[02:50:47] <erdgeist> alus: most often packets don't grow significantly by adding another info_hash
[02:52:24] <erdgeist> btw.. does any one have udp success rates? Does this pass through all plastic routing NAT setups? Or through paranoid ISPs?
[02:52:52] <erdgeist> Not taking into account congestion caused packet drops
[02:53:10] <alus> well, Skype works pretty much everywhere
[02:53:25] <The_8472> UDP announces work fine through normal NAT routers
[02:53:30] <erdgeist> they do this via super nodes
[02:53:45] <alus> they use TCP with the super nodes?
[02:53:57] <alus> I don't think so...
[02:54:02] <erdgeist> no, udp. skype is a bad example.
[02:54:14] <The_8472> outgoing UDP usually is not a problem. think DNS.
[02:54:17] <alus> skype is a good example that UDP basically works everywhere
[02:54:25] <erdgeist> you don't have proxies for bittorrent
[02:54:33] <The_8472> fine
[02:54:35] <The_8472> NTP then
[02:54:38] <erdgeist> :)
[02:54:46] <hydri> alus: right.. I think typically it's expressed as: "drops appear in bursts"
[02:55:01] <alus> hydri: right.
[02:55:11] <The_8472> in the absense of AQM and congestion control...
[02:55:29] <alus> hydri: I thought uTP modeled the buffer size in bytes?
[02:55:32] <erdgeist> I was wondering why the tracker protocol was made HTTP in the first place. There MUST have been something to it
[02:55:44] <alus> erdgeist: http is easy to work with
[02:55:47] *** stalled has quit IRC
[02:55:58] <erdgeist> for funny values of "easy"
[02:56:02] <alus> there are lots of servers and web frameworks and load balancers and what have you
[02:56:20] <alus> you could be a static tracker with a flat file and nginx
[02:56:38] <alus> that's a heck of a lot easier than some custom protocol
[02:56:47] <The_8472> think php trackers
[02:56:52] <hydri> alus: I don't think uTP necessarily makes any assumptions about how packet loss occurs, or the buffer size.. you can deduce the buffer size from the rate and max buffer delay I guess
[02:56:52] <alus> and dynamic hosting
[02:56:56] <The_8472> which was pretty much all of the early ones
[02:57:50] <alus> hydri: but the congestion window is in bytes, not packets. oh I see because the time to empty one MTU slot is dependent on the real packet size
[02:57:55] <erdgeist> I thought the early ones were in python
[02:58:02] <hydri> right
[02:58:21] <alus> which means they are not equally size in terms of time... which actually means small packets cause more packet loss than large packets?
[02:58:23] <alus> that's odd.
[02:58:25] <erdgeist> and THAT could've handled custom bencoding streams easily
[02:58:57] <alus> erdgeist: the example was in python. the spec is better than the example
[02:59:07] <erdgeist> it's only.. that http is so overly complex and wasteful
[02:59:17] <alus> complex?
[02:59:25] <The_8472> well, we usually use trimmed down HTTP anyway
[02:59:34] <The_8472> no error code handling, minimalistic headers, ...
[02:59:41] <Nolar> i'd far rather build a massive tracker backend via http than udp...
[02:59:46] <alus> it's not like we're doing chunked encoding
[02:59:54] <erdgeist> http://www.and.org/texts/server-http
[03:00:33] <The_8472> technically we should use POSTs for announces and GET for scrapes and stuff like that... but we don't
[03:00:45] <erdgeist> Yes, if we don't speak http, anyway...
[03:00:47] <The_8472> it's just an RPC protocol under the disguise of HTTP ^^
[03:01:01] <erdgeist> but only pretend to, not follow the syntax ...
[03:01:17] <erdgeist> not follow the grammar...
[03:01:19] <alus> yes, just because it's http compatible doesn't mean you have to support all of HTTP
[03:01:51] <erdgeist> if libcurl turns rogue and actually uses some of what http allows it to do, I'm screwed
[03:02:37] <alus> you should certainly make sure you have control of your library.
[03:02:51] <erdgeist> on which side?
[03:03:03] *** stalled has joined #bittorrent
[03:03:04] <alus> on both sides. don't run software you can't control
[03:04:43] <erdgeist> technically the tracker would have to parse all the funny stuff. My checks for gzip-capabilities of the other end ... is a strstr for gzip in the header
[03:06:17] <alus> the tracker has a better chance at full http support than the clients usually do. but if your client is doing things the tracker doesn't support, obviously that's bad and you should not do that stuff
[03:07:43] <The_8472> erdgeist, you're probably one of the few people not using a library to http.
[03:08:22] *** MassaRoddel has quit IRC
[03:08:50] <alus> The_8472: uT has a custom http client for tracker communication
[03:08:59] <alus> so it's probably not fully compliant
[03:09:26] <erdgeist> Hmm. Thats not the point. The point still is, that http is overly complex for the job at hand, introduces some funny layering issues and has an ugly overhead
[03:10:04] <The_8472> well, historic reasons. nobody forsaw things like opentracker servicing millions of peers at once
[03:10:35] <The_8472> and the UDP protocol is just as ugly ^^
[03:10:50] <The_8472> in the other direction... not overly complex... it's too unflexible
[03:10:57] <erdgeist> didn't utorrent1600 also make up its own rquest string and nearly killed the tpb tracker?
[03:11:10] <erdgeist> ys, I dont like the udp proto either
[03:11:28] <erdgeist> But I was young and stuck to the only proposal I found, instead of making it better
[03:11:54] <alus> just replace the packet format with bencode. perhaps that's all you need
[03:12:06] <alus> the rest is just semantics
[03:12:09] <erdgeist> yes, that would have been smarter
[03:12:25] <The_8472> works reasonably well for the DHT
[03:12:34] <The_8472> though we occasionally run into problems with packet sizes ^^
[03:12:41] <erdgeist> or at least send http over udp to reuse the parser
[03:12:47] <erdgeist> eheh
[03:13:10] <The_8472> http over udp... we already have to do that for upnp
[03:13:52] <The_8472> which is a soul sucking monstrosity by itself
[03:13:53] <erdgeist> at least it saves the tcp overhead and still php can handle it
[03:14:07] <erdgeist> didn't microsoft invent this?
[03:14:17] <The_8472> cgi php? ^^
[03:14:18] <erdgeist> or was it nat-pmp?
[03:14:34] <The_8472> nat-pmp is apple i think
[03:16:07] <erdgeist> but isn't there a java class for http over udp?
[03:17:07] <Nolar> heh, no, but probably not hard to do
[03:17:22] <Nolar> but not really part of any std lib
[03:17:37] <erdgeist> ohh, already 3am, again. Time to sleep.
[03:17:39] <The_8472> you could probably use the one from apache
[03:17:48] <The_8472> and tack on udp support
[03:18:12] <The_8472> sun's urlhttpconnection is ugly anyway. 0 control over the underlying sockets
[03:18:15] <erdgeist> and considering the amount of beer bottles in front of me it's time to adjust the alarm to +1h :)
[03:18:25] <The_8472> hahaha
[03:18:29] <Nolar> ya, just need to get some http looking output to stuff in a packet
[03:19:27] <The_8472> can't even bind the source address/port of http connections with sun's. much to my annoyance
[03:20:48] <erdgeist> Well. We could just decide upon using udpV2.
[03:21:46] *** mxs has quit IRC
[03:21:49] *** mxs has joined #bittorrent
[03:26:51] *** The_8472 has quit IRC
[03:27:12] *** wadim has joined #bittorrent
[03:27:14] *** wadim is now known as The_8472
[03:33:06] <alus> The_8472: do we run in to packet size issues in the DHT due to bencode or due to lots of data to send? what's the maximum overhead of bencode in a DHT packet?
[03:39:06] <The_8472> lots of data to send
[04:05:31] *** The_8472 has quit IRC
[04:05:52] *** wadim has joined #bittorrent
[04:05:54] *** wadim is now known as The_8472
[04:27:10] *** Nolar_ has joined #bittorrent
[04:27:48] *** Nolar has quit IRC
[04:28:19] *** Nolar_ is now known as Nolar
[04:32:03] *** Nolar has quit IRC
[04:32:03] *** goussx has quit IRC
[04:32:22] *** goussx has joined #bittorrent
[04:36:52] *** goussx has quit IRC
[04:37:55] *** goussx has joined #bittorrent
[04:40:21] *** Nolar has joined #bittorrent
[04:54:46] *** goussx has quit IRC
[04:59:07] *** chelz has joined #bittorrent
[05:02:20] *** goussx has joined #bittorrent
[06:20:21] *** goussx has quit IRC
[06:58:02] *** goussx has joined #bittorrent
[06:58:16] *** goussx_ has joined #bittorrent
[07:14:44] *** goussx__ has joined #bittorrent
[07:15:38] *** goussx__ has quit IRC
[07:15:51] *** Nolar has quit IRC
[07:15:51] *** goussx has quit IRC
[07:15:58] *** goussx__ has joined #bittorrent
[07:16:09] *** goussx__ is now known as goussx
[07:18:30] *** Nolar has joined #bittorrent
[07:18:58] *** MassaRoddel has joined #bittorrent
[07:24:27] *** _rafi_ has joined #bittorrent
[07:32:05] *** goussx_ has quit IRC
[08:08:42] *** stalled_ has joined #bittorrent
[08:19:48] *** stalled has quit IRC
[08:19:49] *** stalled_ is now known as stalled
[08:25:54] *** chelz has quit IRC
[08:51:47] *** goussx has quit IRC
[09:09:11] *** chelz has joined #bittorrent
[09:27:56] *** goussx has joined #bittorrent
[10:07:31] *** init0_ has joined #bittorrent
[10:18:16] *** init0 has quit IRC
[10:37:03] *** kwinz2 has quit IRC
[10:37:22] *** kwinz2 has joined #bittorrent
[11:02:55] *** goussx has quit IRC
[11:16:48] *** ivan`` has joined #bittorrent
[11:25:49] *** ivan` has quit IRC
[11:33:35] *** chelz has quit IRC
[11:37:57] *** kwinz2 has quit IRC
[11:49:33] *** goussx has joined #bittorrent
[12:31:53] *** kwinz2 has joined #bittorrent
[12:39:13] *** DreadWingKnight has quit IRC
[12:56:00] <Koper> What should the tracker answer when it recives a "completed" event ?
[12:57:22] *** goussx_ has joined #bittorrent
[13:00:03] <Koper> err I meant
[13:00:08] <Koper> when it receives a "stopped" event
[13:00:16] <Koper> doesn't make much sense to answer with a list of peers..
[13:06:29] *** goussx has quit IRC
[13:06:29] *** goussx_ is now known as goussx
[13:13:43] *** DWKnight has joined #bittorrent
[13:23:07] <alus> Koper: completed just means the client is done downloading
[13:23:39] <alus> Koper: it will probably continue uploading, so additional peers (who are downloading) may be useful
[13:27:05] <DWKnight> numwant=0 and event=stopped are the only reasons to not send a peerlist in an announce reply
[13:28:48] <Koper> I meant to type stopped alus, not completed
[13:29:03] <Koper> DWKnight: yes, but in that case what do I send? An empty text/plain ?
[14:22:26] *** ivan`` is now known as ivan`
[14:22:55] *** ivan` is now known as Guest6008
[14:29:00] *** mxs has quit IRC
[14:29:00] *** Astro has quit IRC
[14:29:01] *** erdgeist has quit IRC
[14:29:01] *** jnpplf has quit IRC
[14:29:01] *** GTHK has quit IRC
[14:29:01] *** BitTorrentBot has quit IRC
[14:29:02] *** alus has quit IRC
[14:29:02] *** charles has quit IRC
[14:29:02] *** tris has quit IRC
[14:29:03] *** swinokur has quit IRC
[14:29:04] *** init0_ has quit IRC
[14:29:04] *** The_8472 has quit IRC
[14:29:05] *** A9[idle] has quit IRC
[14:29:05] *** wojci has quit IRC
[14:29:05] *** mpl has quit IRC
[14:29:05] *** Koper has quit IRC
[14:29:05] *** r3wj has quit IRC
[14:29:05] *** deltab has quit IRC
[14:29:05] *** hlindhe_ has quit IRC
[14:29:05] *** Firon has quit IRC
[14:29:06] *** burris has quit IRC
[14:29:06] *** uriel has quit IRC
[14:29:06] *** K`Tetch has quit IRC
[14:29:07] *** _Mazon has quit IRC
[14:29:07] *** goussx has quit IRC
[14:29:07] *** kwinz2 has quit IRC
[14:29:08] *** ProperNoun has quit IRC
[14:29:08] *** HandheldPenguin` has quit IRC
[14:29:08] *** TheSHAD0W has quit IRC
[14:29:08] *** DWKnight has quit IRC
[14:29:08] *** _rafi_ has quit IRC
[14:29:09] *** BentMyWookie has quit IRC
[14:29:09] *** alienvenom has quit IRC
[14:29:09] *** cgreco has quit IRC
[14:29:10] *** hlindhe has quit IRC
[14:29:10] *** Gottaname has quit IRC
[14:29:10] *** DeHackEd has quit IRC
[14:29:11] *** swolchok has quit IRC
[14:29:11] *** init0_ has joined #bittorrent
[14:29:11] *** The_8472 has joined #bittorrent
[14:29:11] *** swinokur has joined #bittorrent
[14:29:11] *** r3wj has joined #bittorrent
[14:29:11] *** A9[idle] has joined #bittorrent
[14:29:12] *** deltab has joined #bittorrent
[14:29:12] *** wojci has joined #bittorrent
[14:29:12] *** mpl has joined #bittorrent
[14:29:12] *** Koper has joined #bittorrent
[14:29:12] *** Firon has joined #bittorrent
[14:29:12] *** hlindhe has joined #bittorrent
[14:29:30] *** goussx has joined #bittorrent
[14:29:30] *** kwinz2 has joined #bittorrent
[14:29:30] *** TheSHAD0W has joined #bittorrent
[14:29:30] *** K`Tetch has joined #bittorrent
[14:29:30] *** _Mazon has joined #bittorrent
[14:29:30] *** uriel has joined #bittorrent
[14:29:30] *** HandheldPenguin` has joined #bittorrent
[14:29:30] *** burris has joined #bittorrent
[14:29:30] *** ProperNoun has joined #bittorrent
[14:29:51] *** GTHK has joined #bittorrent
[14:30:57] *** cgreco has joined #bittorrent
[14:31:10] *** charles has joined #bittorrent
[14:31:11] *** DWKnight has joined #bittorrent
[14:31:11] *** _rafi_ has joined #bittorrent
[14:31:11] *** BentMyWookie has joined #bittorrent
[14:31:11] *** alienvenom has joined #bittorrent
[14:31:47] *** hlindhe has quit IRC
[14:31:50] *** hlindhe_ has joined #bittorrent
[14:31:51] *** Gottaname has joined #bittorrent
[14:31:51] *** hlindhe has joined #bittorrent
[14:31:51] *** DeHackEd has joined #bittorrent
[14:31:51] *** swolchok has joined #bittorrent
[14:32:33] *** TheSHAD0W has quit IRC
[14:33:34] *** jnpplf has joined #bittorrent
[14:33:34] *** BitTorrentBot has joined #bittorrent
[14:34:01] *** erdgeist has joined #bittorrent
[14:34:14] *** alus has joined #bittorrent
[14:39:58] *** Astro has joined #bittorrent
[14:43:49] *** stalled has quit IRC
[14:46:37] *** mxs has joined #bittorrent
[14:49:57] *** stalled has joined #bittorrent
[14:51:08] <DWKnight> Koper: normal reply with a 0-length peers value
[14:54:49] *** tris has joined #bittorrent
[15:09:49] <DWKnight> also, netsplits suck
[15:21:04] *** Guest6008 is now known as ivan`
[15:21:29] *** ivan` is now known as Guest19757
[15:29:18] *** DeHackEd has quit IRC
[15:29:18] *** hlindhe has quit IRC
[15:29:18] *** swolchok has quit IRC
[15:29:20] *** Gottaname has quit IRC
[15:29:21] *** _rafi_ has quit IRC
[15:29:21] *** DWKnight has quit IRC
[15:29:21] *** BentMyWookie has quit IRC
[15:29:21] *** alienvenom has quit IRC
[15:29:22] *** mxs has quit IRC
[15:29:22] *** Astro has quit IRC
[15:29:22] *** jnpplf has quit IRC
[15:29:22] *** BitTorrentBot has quit IRC
[15:29:23] *** cgreco has quit IRC
[15:29:23] *** burris has quit IRC
[15:29:24] *** _Mazon has quit IRC
[15:29:24] *** uriel has quit IRC
[15:29:24] *** K`Tetch has quit IRC
[15:29:24] *** tris has quit IRC
[15:29:24] *** charles has quit IRC
[15:29:25] *** kwinz2 has quit IRC
[15:29:25] *** goussx has quit IRC
[15:29:25] *** ProperNoun has quit IRC
[15:29:25] *** HandheldPenguin` has quit IRC
[15:29:26] *** stalled has quit IRC
[15:29:26] *** alus has quit IRC
[15:29:26] *** hlindhe_ has quit IRC
[15:29:26] *** GTHK has quit IRC
[15:29:26] *** swinokur has quit IRC
[15:29:27] *** erdgeist has quit IRC
[15:29:29] *** Koper has quit IRC
[15:29:29] *** mpl has quit IRC
[15:29:29] *** A9[idle] has quit IRC
[15:29:29] *** The_8472 has quit IRC
[15:29:29] *** init0_ has quit IRC
[15:29:29] *** Firon has quit IRC
[15:29:29] *** wojci has quit IRC
[15:29:29] *** deltab has quit IRC
[15:29:29] *** r3wj has quit IRC
[15:29:30] *** Guest19757 has quit IRC
[15:29:30] *** MassaRoddel has quit IRC
[15:29:32] *** andar has quit IRC
[15:29:32] *** kjetilho has quit IRC
[15:29:32] *** foomor has quit IRC
[15:29:32] *** bittwist has quit IRC
[15:29:33] *** nks has quit IRC
[15:29:33] *** void^ has quit IRC
[15:29:34] *** uau has quit IRC
[15:29:35] *** tris has joined #bittorrent
[15:29:35] *** stalled has joined #bittorrent
[15:29:35] *** mxs has joined #bittorrent
[15:29:35] *** Astro has joined #bittorrent
[15:29:35] *** alus has joined #bittorrent
[15:29:35] *** erdgeist has joined #bittorrent
[15:29:35] *** BitTorrentBot has joined #bittorrent
[15:29:35] *** jnpplf has joined #bittorrent
[15:29:35] *** swolchok has joined #bittorrent
[15:29:35] *** DeHackEd has joined #bittorrent
[15:29:35] *** hlindhe has joined #bittorrent
[15:29:35] *** Gottaname has joined #bittorrent
[15:29:35] *** hlindhe_ has joined #bittorrent
[15:29:35] *** alienvenom has joined #bittorrent
[15:29:35] *** BentMyWookie has joined #bittorrent
[15:29:35] *** _rafi_ has joined #bittorrent
[15:29:35] *** DWKnight has joined #bittorrent
[15:29:35] *** charles has joined #bittorrent
[15:29:35] *** cgreco has joined #bittorrent
[15:29:35] *** GTHK has joined #bittorrent
[15:29:35] *** ProperNoun has joined #bittorrent
[15:29:35] *** burris has joined #bittorrent
[15:29:35] *** HandheldPenguin` has joined #bittorrent
[15:29:35] *** uriel has joined #bittorrent
[15:29:35] *** _Mazon has joined #bittorrent
[15:29:35] *** K`Tetch has joined #bittorrent
[15:29:35] *** kwinz2 has joined #bittorrent
[15:29:35] *** goussx has joined #bittorrent
[15:29:35] *** Firon has joined #bittorrent
[15:29:35] *** Koper has joined #bittorrent
[15:29:35] *** mpl has joined #bittorrent
[15:29:35] *** wojci has joined #bittorrent
[15:29:35] *** deltab has joined #bittorrent
[15:29:35] *** A9[idle] has joined #bittorrent
[15:29:35] *** r3wj has joined #bittorrent
[15:29:35] *** swinokur has joined #bittorrent
[15:29:35] *** The_8472 has joined #bittorrent
[15:29:35] *** init0_ has joined #bittorrent
[15:29:35] *** Guest19757 has joined #bittorrent
[15:29:35] *** MassaRoddel has joined #bittorrent
[15:29:35] *** uau has joined #bittorrent
[15:29:35] *** bittwist has joined #bittorrent
[15:29:35] *** kjetilho has joined #bittorrent
[15:29:35] *** foomor has joined #bittorrent
[15:29:35] *** nks has joined #bittorrent
[15:29:35] *** andar has joined #bittorrent
[15:29:35] *** void^ has joined #bittorrent
[15:29:46] *** hydri_ has joined #bittorrent
[15:30:36] *** PN has joined #bittorrent
[15:31:07] *** stalled has quit IRC
[15:32:45] *** hlindhe has quit IRC
[15:32:45] *** Gottaname has quit IRC
[15:32:45] *** DeHackEd has quit IRC
[15:32:45] *** swolchok has quit IRC
[15:32:47] *** _rafi_ has quit IRC
[15:32:47] *** BentMyWookie has quit IRC
[15:32:47] *** DWKnight has quit IRC
[15:32:47] *** alienvenom has quit IRC
[15:32:48] *** Astro has quit IRC
[15:32:48] *** mxs has quit IRC
[15:32:48] *** jnpplf has quit IRC
[15:32:49] *** BitTorrentBot has quit IRC
[15:32:49] *** cgreco has quit IRC
[15:32:50] *** burris has quit IRC
[15:32:50] *** _Mazon has quit IRC
[15:32:50] *** uriel has quit IRC
[15:32:50] *** K`Tetch has quit IRC
[15:32:51] *** goussx has quit IRC
[15:32:51] *** kwinz2 has quit IRC
[15:32:51] *** HandheldPenguin` has quit IRC
[15:32:52] *** ProperNoun has quit IRC
[15:32:52] *** charles has quit IRC
[15:32:52] *** tris has quit IRC
[15:32:52] *** K`Tetch_ has joined #bittorrent
[15:32:56] *** hlindhe_ is now known as hlindhe
[15:33:25] *** stalled_ has joined #bittorrent
[15:33:25] *** mxs has joined #bittorrent
[15:33:25] *** Astro has joined #bittorrent
[15:33:25] *** BitTorrentBot has joined #bittorrent
[15:33:25] *** jnpplf has joined #bittorrent
[15:33:25] *** cgreco has joined #bittorrent
[15:33:42] *** DWKnight has joined #bittorrent
[15:33:42] *** _rafi_ has joined #bittorrent
[15:33:42] *** BentMyWookie has joined #bittorrent
[15:33:42] *** alienvenom has joined #bittorrent
[15:34:22] *** stalled_ is now known as stalled
[15:34:31] *** tris has joined #bittorrent
[15:34:31] *** charles has joined #bittorrent
[15:34:31] *** goussx has joined #bittorrent
[15:34:31] *** kwinz2 has joined #bittorrent
[15:34:31] *** _Mazon has joined #bittorrent
[15:34:31] *** uriel has joined #bittorrent
[15:34:31] *** HandheldPenguin` has joined #bittorrent
[15:34:31] *** burris has joined #bittorrent
[15:34:31] *** ProperNoun has joined #bittorrent
[15:37:16] *** Waldorf has joined #bittorrent
[15:38:38] *** Switeck has joined #bittorrent
[15:45:03] *** hydri has quit IRC
[15:45:03] *** hydri_ is now known as hydri
[15:46:44] *** burris has quit IRC
[15:46:44] *** _Mazon has quit IRC
[15:46:44] *** uriel has quit IRC
[15:47:00] *** burris has joined #bittorrent
[15:47:01] *** Mazon has joined #BitTorrent
[15:49:10] *** hlindhe has quit IRC
[15:49:13] *** hlindhe_ has joined #bittorrent
[15:49:14] *** Gottaname has joined #bittorrent
[15:49:14] *** hlindhe has joined #bittorrent
[15:49:14] *** DeHackEd has joined #bittorrent
[15:49:14] *** swolchok has joined #bittorrent
[15:51:57] *** uriel has joined #bittorrent
[15:54:02] *** stalled has quit IRC
[15:54:39] *** Guest19757 is now known as ivan`
[15:55:09] *** ivan` is now known as Guest22617
[15:56:09] *** K`Tetch has joined #bittorrent
[16:01:52] *** ProperNoun has quit IRC
[16:01:58] *** stalled has joined #bittorrent
[16:05:37] *** K`Tetch_ has quit IRC
[16:22:19] *** init0_ has quit IRC
[16:51:57] *** Guest22617 is now known as ivan`
[16:52:26] *** ivan` is now known as Guest64146
[16:52:28] *** foomor has quit IRC
[16:52:51] <Switeck> udp://tracker.publicbt.com:80/announce  seems to be returning a list of bogus ip addresses. Anyone else having the same problem?
[16:55:15] <The_8472> have you tried with differnet clients?
[16:55:22] <The_8472> maybe the parsing is off
[16:55:32] *** Guest64146 is now known as ivan`
[16:55:33] <Switeck> that's why I'm asking here 8)
[16:56:13] <Switeck> I know that url is often used...I've see the same thing happen with the udp tracker for tracker.openbittorrent.com as well
[16:56:36] <The_8472> well, not sure... i wouldn't notice, if the tracker returns crap dht will just pick up the job ^^
[16:57:04] <The_8472> i guess a torrent marked as private running on those trackers would provide more insight
[16:57:14] <charles> Switeck: what client/version?  uTorrent?
[16:57:18] <Switeck> yes
[16:57:36] <Switeck> I have some torrents whose entire torrent swarm are almost certainly <10 seeds+peers, yet due to the udp tracker I'm getting 50+ ips...often in ridiculous ranges.
[16:58:14] <The_8472> well, could be a problem on either end
[16:58:25] 
[16:59:40] <Switeck> I'm reluctant to attempt to troubleshoot every possibility. :P
[17:04:02] *** foomor has joined #bittorrent
[17:04:40] *** ajaya has joined #bittorrent
[17:13:56] *** burris has quit IRC
[18:20:48] *** ajaya has quit IRC
[18:25:21] *** deltab has quit IRC
[18:27:56] *** goussx has quit IRC
[18:39:51] *** neonfloor has joined #bittorrent
[18:40:06] *** Waldorf has quit IRC
[18:41:13] *** ahath has quit IRC
[18:41:56] *** ajaya has joined #bittorrent
[18:49:27] *** charles_ has joined #bittorrent
[19:04:05] *** adac has joined #bittorrent
[19:04:30] <adac> Can someone tell me why it is impossible to connect to all sources that are available?
[19:05:06] <DWKnight> first up
[19:05:07] <DWKnight> firewalls
[19:05:10] *** goussx has joined #bittorrent
[19:05:12] <DWKnight> on their end and/or yours
[19:05:17] <DWKnight> second up
[19:05:22] <Switeck> Second: reports are exaggerated.
[19:05:26] <DWKnight> not all sources you think are available actually are
[19:05:28] <DWKnight> third up
[19:05:40] <DWKnight> it's fucking stupid to even try once the swarm size exceeds about 100 peers
[19:05:56] <Switeck> I second that last statement
[19:06:33] <adac> aorry, last one I do not understand
[19:07:00] <Switeck> Simple: you can't upload to that many people effectively with much less than a 100 mbit/sec upload line
[19:07:18] <Switeck> So attempting to download from so many implies outright leeching from many.
[19:07:55] <Switeck> If you're getting terrible speeds from 10 peers/seeds, increasing to 100 peers/seeds may even decrease your speeds.
[19:08:13] <adac> Ok I see
[19:39:19] *** Waldorf has joined #bittorrent
[20:14:18] *** kwinz2 has quit IRC
[20:16:06] *** sobczyk_ has joined #bittorrent
[20:16:14] <sobczyk_> hi, I have a question loosely connected with rtotrrent but torrent protocol in general. Does BT optimize or oragnize network connections? (excluding DHT)
[20:16:43] <sobczyk_> uhh just cpy pated the question from other channel :0
[21:02:23] *** echelog has joined #bittorrent
[21:02:41] *** Mazon has joined #BitTorrent
[21:05:57] *** bittwist has quit IRC
[21:16:26] *** KyleK_ has joined #bittorrent
[21:19:26] *** edigaryev has joined #bittorrent
[22:16:33] *** foomor has left #bittorrent
[22:24:35] *** sobczyk_ is now known as sobczyk
[22:25:57] *** kwinz2 has joined #bittorrent
[22:30:42] *** edigaryev has quit IRC
[22:45:17] *** andar2 has joined #bittorrent
[23:18:14] *** adac has quit IRC
[23:20:17] *** init0 has joined #bittorrent
[23:27:24] *** Nolar has quit IRC
[23:27:51] *** Nolar has joined #bittorrent
[23:46:21] *** neitcho has joined #bittorrent
[23:53:06] *** KyleK_ has quit IRC





top