[FFmpeg-trac] #5080(undetermined:new): Excessive HTTP GETs reading MP4 from web server

[FFmpeg]

#5080: Excessive HTTP GETs reading MP4 from web server
-------------------------------------+-------------------------------------
             Reporter:  nealzebub    |                    Owner:
                 Type:  defect       |                   Status:  new
             Priority:  normal       |                Component:
              Version:  git-master   |  undetermined
             Keywords:  http mov     |               Resolution:
             Blocking:               |               Blocked By:
Analyzed by developer:  0            |  Reproduced by developer:  0
-------------------------------------+-------------------------------------

Comment (by stefang):

 You can see from the list of A/V positions that the file has a drift in
 multiplexing. By that I mean that while at the beginning of the list the
 difference in position between audio and video chunks of the same
 timestamp are about 6k (audio with pts 0 at pos 44, video with pts 0 at
 pos 6210) this increases towards the end of the list to about 30k (audio
 with pts 34.3 at pos 647105, video with pts 34.3 at pos 675027).
 I have also seen such files in the wild on numerous occasions but they are
 hard to handle nicely. For smaller files as in your case just increasing
 the size of some I/O buffers might help. But for movies of hours length
 which drift apart by many megabytes the currently employed solution is not
 so bad, handling it without additional GETs would require complex code and
 would increase buffer size requirements by a lot.
 On the content side, such files can be improved by remultiplexing or using
 http://linux.die.net/man/3/mp4optimize

--
Ticket URL: <https://trac.ffmpeg.org/ticket/5080#comment:6>
FFmpeg <https://ffmpeg.org>
FFmpeg issue tracker