Re: [FFmpeg-devel] [GSoC] FFserver: Add HLS and DASH
On Sun, Jul 8, 2018 at 11:21 PM, Michael Niedermayer wrote: > On Thu, Jun 28, 2018 at 02:51:00AM +0200, Stephan Holljes wrote: >> This patchset took embarrassingly long to produce. I first took a few wrong >> turns, but I think I'm now at a point where I reached the limits of the >> public server API in libavformat. The main problem is that I cannot set >> options on clients that are accepted through http_accept(), for example >> a timeout on socket operations, as currently a client that disconnects while >> being sent a file causes the sending thread to loop infintely in a >> poll()-call. >> I wonder if adding a different library handling the http server (as planned >> for >> the future) would be better done now rather than later. >> >> Another bug I cannot really explain so far (that may be because of my >> unfamiliarity with DASH) is that with longer files the DASH stream starts to >> repeat some fragments (at least in mpv) in weird ways. HLS does not have this >> problem. >> >> Another thing that is not yet handled are unmuxable codecs. >> >> The patchset also includes some cleanup and other various fixes. > > on a slightly different subject, do people want me to create a git repository > on git.ffmpeg.org, similar to: > > https://git.ffmpeg.org/gitweb/ffmpeg.git > > for this ? > as tha patches are not for the main ffmpeg repository ... > or maybe ive missed a discussion about this and this already has a intended > place to be put to ? I think it was briefly discussed, but not properly communicated. As far as I know a separate repository was the optimal solution. > > if i should create one, who should have access?, who will maintain it? > This question about maintaince is probably somewhat for mentor and student, > as i assume you know the code much better than anyone else ... My first reaction is that I am up to maintaining it and I hope that I am up to the task. Of course help from other people that are experienced in network stuff is greatly appreciated. I haven't been available much the past few days on IRC so I haven't discussed this with anyone else. > > thanks > > [...] > -- > Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB > > Complexity theory is the science of finding the exact solution to an > approximation. Benchmarking OTOH is finding an approximation of the exact > > ___ > ffmpeg-devel mailing list > ffmpeg-devel@ffmpeg.org > http://ffmpeg.org/mailman/listinfo/ffmpeg-devel > ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [GSoC] FFserver: Add HLS and DASH
On Thu, Jun 28, 2018 at 02:51:00AM +0200, Stephan Holljes wrote: > This patchset took embarrassingly long to produce. I first took a few wrong > turns, but I think I'm now at a point where I reached the limits of the > public server API in libavformat. The main problem is that I cannot set > options on clients that are accepted through http_accept(), for example > a timeout on socket operations, as currently a client that disconnects while > being sent a file causes the sending thread to loop infintely in a > poll()-call. > I wonder if adding a different library handling the http server (as planned > for > the future) would be better done now rather than later. > > Another bug I cannot really explain so far (that may be because of my > unfamiliarity with DASH) is that with longer files the DASH stream starts to > repeat some fragments (at least in mpv) in weird ways. HLS does not have this > problem. > > Another thing that is not yet handled are unmuxable codecs. > > The patchset also includes some cleanup and other various fixes. on a slightly different subject, do people want me to create a git repository on git.ffmpeg.org, similar to: https://git.ffmpeg.org/gitweb/ffmpeg.git for this ? as tha patches are not for the main ffmpeg repository ... or maybe ive missed a discussion about this and this already has a intended place to be put to ? if i should create one, who should have access?, who will maintain it? This question about maintaince is probably somewhat for mentor and student, as i assume you know the code much better than anyone else ... thanks [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB Complexity theory is the science of finding the exact solution to an approximation. Benchmarking OTOH is finding an approximation of the exact signature.asc Description: PGP signature ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [GSoC] FFserver: Add HLS and DASH
On Thu, Jun 28, 2018 at 2:51 AM, Stephan Holljes wrote: > This patchset took embarrassingly long to produce. I first took a few wrong > turns, but I think I'm now at a point where I reached the limits of the > public server API in libavformat. The main problem is that I cannot set > options on clients that are accepted through http_accept(), for example > a timeout on socket operations, as currently a client that disconnects while > being sent a file causes the sending thread to loop infintely in a > poll()-call. > I wonder if adding a different library handling the http server (as planned > for > the future) would be better done now rather than later. > > Another bug I cannot really explain so far (that may be because of my > unfamiliarity with DASH) is that with longer files the DASH stream starts to > repeat some fragments (at least in mpv) in weird ways. HLS does not have this > problem. > > Another thing that is not yet handled are unmuxable codecs. > > The patchset also includes some cleanup and other various fixes. > > Ping, since I sent this email I've been working on adding libmicrohttpd to replace lavformat's http-server. ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
[FFmpeg-devel] [GSoC] FFserver: Add HLS and DASH
This patchset took embarrassingly long to produce. I first took a few wrong turns, but I think I'm now at a point where I reached the limits of the public server API in libavformat. The main problem is that I cannot set options on clients that are accepted through http_accept(), for example a timeout on socket operations, as currently a client that disconnects while being sent a file causes the sending thread to loop infintely in a poll()-call. I wonder if adding a different library handling the http server (as planned for the future) would be better done now rather than later. Another bug I cannot really explain so far (that may be because of my unfamiliarity with DASH) is that with longer files the DASH stream starts to repeat some fragments (at least in mpv) in weird ways. HLS does not have this problem. Another thing that is not yet handled are unmuxable codecs. The patchset also includes some cleanup and other various fixes. ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel