On Fri, Oct 25, 2002 at 03:42:53AM +1000, fish wrote: > > as usual, linux: > http://www.dspaudio.com/~jaymz/fishstream-fec-linux.tar.gz > > win32: > http://www.dspaudio.com/~jaymz/streamserver.zip > http://www.dspaudio.com/~jaymz/freenet-encoder.zip > > So, this is the mythical test of using FEC encoding with 50% > redundancy. i've included a binary of the new FEC module, however, you > might have to build it yourself - i stole the code from nobody else, who > is clearly much better at the whole module thing than I am :) > > the new test stream is at > http://localhost:8892/SSK at kWu5Osv~VAI3-kH7z8QIVxklv-YPAgM/bgc2040-fec.ogg > , however, it was only inserted at HTL=5, which means that it's patchy at Why not use HTL=10? Many freesites are inserted at this HTL. Maybe even 15, for this particular application. > best, for now (freenet seems ratehr fond of dropping data on the floor > still, tho obviously less so now that we're DSB-Free(tm)) - on a clean > transient node, I was only able to retrieve the first 4 minutes on my > first attempt, and 7 on my second, before the DNF's set in. Better then > I'd have done without FEC, but still, kinda sucks. > > However, things get a lot less pessemistic if you're running a permanent > node. it seems that transient nodes are often nigh on useless when it > comes to routing..... and the fact that things get better over time means > that perhaps I should increase the number of retries (at 3, right now, > with htl=htl+5 at the start of each), tho it's possible that this is Yeah, maybe... thanks to random first hop, more retries may make sense. But a higher insertion htl will make a bigger impact. > perhaps a side effect of DNF caching (which i know makes sense in the > context of the network blah blah, yeah, i'm aware of it's good points, not > flaming) > > side note: i used to have a lot of success running a permanode behind a > nat... i found it routed *much* better than a transient did iun hte same > config. weird. But I havn't tried this for ages.... > > Anyhow, about the FEC, currently, it's implented by using t second chunks, > each FEC encoded into n/k chunks. the defaults for these t=60, n=6, k=4, > which provides 50% redundancy on each minute of audio, and happens to be a > good size that i don't need to use redirects to fit into keys (around 30k > per fec block) > > oh, yeah, connecting and getting the stream started still takes *far* too > long for my liking. I'm working on that... that's another one of those > things that (a) inserting things at higher htl's and (b) being a permanode > both help > > - fish > > > _______________________________________________ > devl mailing list > devl at freenetproject.org > http://hawk.freenetproject.org/cgi-bin/mailman/listinfo/devl >
-- Matthew Toseland toad at amphibian.dyndns.org amphibian at users.sourceforge.net Freenet/Coldstore open source hacker. Employed full time by Freenet Project Inc. from 11/9/02 to 11/11/02. http://freenetproject.org/ -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 189 bytes Desc: not available URL: <https://emu.freenetproject.org/pipermail/devl/attachments/20021024/4e5e1c92/attachment.pgp>
