[MPlayer-users] Incomplete file behaviour

[Reimar Döffinger]

On Tue, Apr 01, 2008 at 12:05:51AM +0200, belcampo wrote:
> Reimar Döffinger wrote:
> > On Mon, Mar 31, 2008 at 09:06:00PM +0200, belcampo wrote:
> >> Reimar Döffinger wrote:
> >>> On Mon, Mar 31, 2008 at 03:38:15PM +0200, belcampo wrote:
> >>>> If the contents of a 1000 sec h264 file would be on a distributed 
> >>>> filesystem, on say a 1000 computers, and 1 would not be available, so 25 
> >>>> frames would be missing, how would mplayer react on this, hang, or skip 
> >>>> 1 sec and play the rest of the file ?
> >>> Sorry, this is too little information to even _guess_. What do you
> >>> consider a "h264 file"? H.264 Annex B / ES or H.264 in MPEG-TS, or in
> >>> MPEG-PS or in MOV/MP4?
> >> MOV/MP4
> > 
> > Since MOV/MP4 _needs_ and index, one missing piece (at beginning or end)
> > can mean that the file is impossible to play, no matter which player you
> > use.
> It's a hinted MP4. I can start playing a remote-file because file-info 
> is at the beginning

That doesn't change the fact that if the first part of the file is
missing it is completely unplayable. MOV/MP4 is the worst format by far to
use on a unreliable media.

> >>>   does it just return zeros, random data, or create and I/O error?
> >> create I/O error
> > 
> > MPlayer will stop playing if it encounters an I/O error. This could be
> > changed (by changing stream/stream_file.c the fill_buffer function to
> > e.g. return a buffer filled with all 0 when the read fails), though the
> > problem is that an I/O error is too little information, how do you know
> > which is the next part of the file that is available? 
> I think/hope/believe it will be in the mp4 container

How on earth should the MP4 container know which parts of the file are
broken due to some computers being offline?? I guess whatever you're
trying to do is quite feasible, but it is not simple and I think it is
misusing distributed file systems for something they were never designed
for.