*** pinc|mirror has joined the channel2009-12-13 14:03:04
*** pinc has left the channel2009-12-13 14:03:04
*** scott____ has joined the channel2009-12-13 14:32:23
*** Vader has joined the channel2009-12-13 14:55:13
*** Guest4716033 has left the channel2009-12-13 14:56:56
*** scott____ has left the channel2009-12-13 15:44:45
*** mike_____ has joined the channel2009-12-13 16:13:24
*** pinc|mirror has left the channel2009-12-13 17:16:28
*** pinc has joined the channel2009-12-13 18:17:17
*** pinc has left the channel2009-12-13 18:17:24
*** Shelwien has left the channel2009-12-13 18:28:38
*** pinc has joined the channel2009-12-13 22:08:36
*** Krugz has joined the channel2009-12-13 22:20:33
*** Shelwien has joined the channel2009-12-13 22:38:15
*** pinc has left the channel2009-12-13 22:47:33
*** mike_____ has left the channel2009-12-13 23:45:24
*** osman has joined the channel2009-12-14 00:15:03
<osman> hello all2009-12-14 00:15:13
<Shelwien> hi ;)2009-12-14 00:15:53
<osman> how are you?2009-12-14 00:16:30
<Shelwien> ok i guess ;) and you? ;)2009-12-14 00:17:17
<osman> a bit tired and sleepy as usual ;)2009-12-14 00:18:23
 i'm trying to catch long-time delay of icq log :)2009-12-14 00:18:46
 it's too long2009-12-14 00:18:51
<Shelwien> well, i posted an "archiver" recently2009-12-14 00:19:03
 !grep scan72009-12-14 00:19:06
<osman> good :)2009-12-14 00:20:13
<Shelwien> and it seems that there's not that much sense in ultra-fast LZ implementations, with that kind of i/o speed ;)2009-12-14 00:20:20
 (and its not like my implementation is slow - rar -m0 speed is similar too)2009-12-14 00:20:57
<osman> if i could reduct some of my "to-do" list items, i'll jump in BIT again 2009-12-14 00:20:59
<Shelwien> how's you DNA btw? ;)2009-12-14 00:21:13
<osman> i only find that even simple delta filter is enough2009-12-14 00:21:31
 but there is only several distinct values.2009-12-14 00:21:44
<Shelwien> ?2009-12-14 00:21:57
<osman> so, then why not to use markov chains for that?2009-12-14 00:22:00
 if you remember i was working on prediction part of which is based on "unwrapped phase linearity"2009-12-14 00:22:32
<Shelwien> yes2009-12-14 00:22:39
<osman> and at last time which have discussed about it, i was trying to use piece-wise linear regression for fitting this linearity2009-12-14 00:23:18
<Shelwien> and?2009-12-14 00:23:39
<osman> and you've offered linear predictor model for practical purpose2009-12-14 00:23:50
 as a first step to jump in linear predictor, i've tried delta filter2009-12-14 00:24:05
 and i got very limited distinct values.2009-12-14 00:24:24
 so, why not to use finite state machines for tracking them?2009-12-14 00:24:44
<Shelwien> i don't really see how to use state machines to track linearity2009-12-14 00:25:13
<osman> imagine, there are only 4 phase angle (usually of course)2009-12-14 00:25:44
<Shelwien> limited set of delta values was kinda expected, knowing the nature of the data2009-12-14 00:25:47
 but you could scale it down a bit2009-12-14 00:26:13
<STalKer-X> are all CM compressors symmetric? o,o2009-12-14 00:26:17
<Shelwien> not necessarily, but most existing ones are2009-12-14 00:26:44
 people that write them are lazy ;)2009-12-14 00:26:58
<osman> so, next phase angle can be known on consequence nucleotids by using a state machine2009-12-14 00:27:14
 btw, Stalker how are you? :)2009-12-14 00:27:32
<STalKer-X> depressed :x2009-12-14 00:27:42
<osman> and who are you? :)2009-12-14 00:27:43
<STalKer-X> hmm :)2009-12-14 00:28:06
<Shelwien> sure, but isn't it the same as directly predicting the next "letter"? 2009-12-14 00:28:15
<osman> of course2009-12-14 00:28:35
 if you remember, i was trying to use it as an extra context2009-12-14 00:28:48
<Shelwien> well, i mentioned that its not easy to make use of that kind of correlations2009-12-14 00:28:56
<osman> just think as SSE2009-12-14 00:29:02
<Shelwien> yeah, i understand of course2009-12-14 00:29:11
<osman> then i have to stick LZ as the others :P2009-12-14 00:29:21
<Shelwien> not really2009-12-14 00:30:03
<osman> anyway, we will see my progress on this. maybe it's totally useless2009-12-14 00:30:14
<Shelwien> and anyway, one way of using it is like this2009-12-14 00:30:39
<osman> there are a few things to do before working on it2009-12-14 00:30:49
<Shelwien> you know that a long sequence of "symbols" are on the same line2009-12-14 00:31:07
 so you can store that information before encoding that sequence2009-12-14 00:31:47
 like line coefs and max "error"2009-12-14 00:32:19
 and that should allow to mask out the "symbols"2009-12-14 00:32:40
<osman> yes, that's another optimization. but, it's only useful on human genome2009-12-14 00:32:50
<Shelwien> which would break out from these limits2009-12-14 00:32:53
<osman> it doesn't work on i.e. e.coli.txt2009-12-14 00:33:04
<Shelwien> well, this approach seems more reasonable to me2009-12-14 00:33:32
 it won't have much effect, but still can provide some additional compression2009-12-14 00:33:59
 and as to SSE2009-12-14 00:34:09
<osman> for e.coli, there is no too much long runs2009-12-14 00:34:15
<Shelwien> well, you know, 1000 would be long enough ;)2009-12-14 00:34:36
 even 100 probably ;)2009-12-14 00:34:43
<osman> as i said before, i'm trying to stick all things together to break some record2009-12-14 00:34:59
<Shelwien> <Shelwien> and as to SSE2009-12-14 00:35:23
<osman> i have to analyze it first for RLE ;)2009-12-14 00:35:32
<Shelwien> i'd try to maintain some kind of average line coefs for recent data2009-12-14 00:36:07
 and maybe use angle or something in SSE context2009-12-14 00:36:27
 but imho that's unlikely to be of much use2009-12-14 00:36:50
 well, unless there're actually some correlations between line coefs and nucleotid probability distributions2009-12-14 00:37:45
<osman> we'll see2009-12-14 00:39:31
 btw, i probably got a job about something like torrent+fma2009-12-14 00:39:52
<Shelwien> %)2009-12-14 00:40:13
<osman> it's from a local TV2009-12-14 00:40:14
<Shelwien> if its for video, there's unlikely to be any use for fma2009-12-14 00:40:54
<osman> it will be determined until weekend2009-12-14 00:41:02
 "something like torrent+fma"2009-12-14 00:41:21
 probably i'll add my "bit stream models"2009-12-14 00:41:45
 and maybe some simple data recovery for incase transmission error 2009-12-14 00:42:16
<Shelwien> can you be more specific? ;)2009-12-14 00:42:19
<osman> also, resume support2009-12-14 00:42:24
<Shelwien> like what is transferred? ;)2009-12-14 00:42:30
<osman> time to time, a huge video file - something like 4-5 GiB2009-12-14 00:42:59
 from one office to the other2009-12-14 00:43:16
<Shelwien> so, as i said, there's no need for fma or compression ;)2009-12-14 00:43:30
<osman> i'll demand a set of sample data ;)2009-12-14 00:44:01
<Shelwien> well, unless you're willing to write a video recompressor ;)2009-12-14 00:44:02
 well, i'd say the best idea should be to use a UDP-based protocol2009-12-14 00:45:26
<osman> my business partner is trying to settle with them. so, i'll learn something later2009-12-14 00:45:32
 yes, i thought that too.2009-12-14 00:45:47
<Shelwien> i mean, transfer the fragments in packets2009-12-14 00:45:59
<osman> and again same idea ;)2009-12-14 00:46:09
<Shelwien> and repeat passes with lost packets, until you collect whole file2009-12-14 00:46:18
<osman> and again same ;)2009-12-14 00:46:32
<Shelwien> otherwise, with tcp and large amounts of data you'd need some safety measures anyway2009-12-14 00:46:51
 and there'd be also tcp overhead2009-12-14 00:47:03
<osman> i expect that they need some specific features, such as only selecting a office name in the list which has a dynamic IP. i plan to solve that via a web service which holds most updated IP list2009-12-14 00:47:57
<Shelwien> in fact, i did implement a weird protocol line that once2009-12-14 00:48:24
 it was a icmp echo tunnel2009-12-14 00:48:31
 it was sending the data as a payload for ping packets2009-12-14 00:49:02
 with fake source ip2009-12-14 00:49:07
<osman> there is no need to make something superb. if it can transfer files in a reasonable time. then it's done :)2009-12-14 00:49:10
<Shelwien> so the targeted machine was "returning" the packets to someplace different from the real sender2009-12-14 00:49:35
 ...and it allowed to download stuff for free from a test account of some provider ;)2009-12-14 00:50:05
 that test account only allowed to access their site2009-12-14 00:50:23
 but with an external server i still could receive "broadcasts" from my server ;)2009-12-14 00:50:42
<osman> :) nice2009-12-14 00:50:53
<Shelwien> i mean, my server sent the "pings" to the provider's server2009-12-14 00:51:13
 and it "returned" them to me ;)2009-12-14 00:51:19
 well, i just remembered about it because there were similar requirements to your case2009-12-14 00:52:09
 so there was a protocol for batch requesting the missing frames after a pass etc2009-12-14 00:52:33
 btw, i'm lazy, so an additional side channel was used for backward transfer (to server) ;)2009-12-14 00:53:18
 i just replaced an image in the forum avatar on provider's site ;)2009-12-14 00:53:57
 and server occasionally checked it ;)2009-12-14 00:54:03
<osman> i'll off now. i have to get up 06:30 %)2009-12-14 01:07:26
<Shelwien> good luck then ;)2009-12-14 01:07:57
<osman> thx2009-12-14 01:08:13
 cya2009-12-14 01:08:15
*** osman has left the channel2009-12-14 01:08:54
*** Shelwien has left the channel2009-12-14 01:13:59
*** Shelwien has joined the channel2009-12-14 01:15:23
*** Krugz|Study has left the channel2009-12-14 01:44:22
*** STalKer-Y has joined the channel2009-12-14 04:50:30
*** STalKer-X has left the channel2009-12-14 04:53:44
*** Krugz|Study has joined the channel2009-12-14 05:22:30
*** Krugz has left the channel2009-12-14 05:52:32
*** pinc has joined the channel2009-12-14 07:41:35
*** mike_____ has joined the channel2009-12-14 14:11:29
*** Krugz has joined the channel2009-12-14 15:18:37
*** pinc has left the channel2009-12-14 16:18:19
<STalKer-Y> 7z changelog is always funny ;D2009-12-14 17:10:37
 "some bugs were fixed" :x2009-12-14 17:10:47
*** pinc has joined the channel2009-12-14 17:15:39
<Krugz> lol should just put "it got better, probably" :P2009-12-14 17:16:05
*** STalKer-Y has left the channel2009-12-14 19:04:53
*** schnaader has joined the channel2009-12-14 19:15:50
*** schnaader has left the channel2009-12-14 19:20:57
*** pinc has left the channel2009-12-14 21:37:42
*** mike_____ has left the channel2009-12-14 23:44:59
*** Krugz has left the channel2009-12-15 00:06:49
*** Krugz has joined the channel2009-12-15 00:14:16
*** Shelwien has left the channel2009-12-15 01:15:11
*** Shelwien has joined the channel2009-12-15 01:51:16
*** STalKer-X has joined the channel2009-12-15 04:49:20
*** pinc has joined the channel2009-12-15 07:13:28
<Shelwien> ...2009-12-15 08:24:41
 !next2009-12-15 08:24:44