[FFmpeg-devel] [PATCH] lavf: add zip protocol

[wm4]

On Sun, 29 Mar 2015 08:55:42 +1100
Peter Ross <pross at xvid.org> wrote:

> On Sat, Mar 28, 2015 at 10:24:55PM +0100, wm4 wrote:
> > On Sun, 29 Mar 2015 08:10:29 +1100
> > Peter Ross <pross at xvid.org> wrote:
> > 
> > > On Sat, Mar 28, 2015 at 08:38:40PM +0100, Lukasz Marek wrote:
> > > > On 28.03.2015 20:13, Nicolas George wrote:
> > > > >L'octidi 8 germinal, an CCXXIII, Lukasz Marek a écrit :
> > > > >>I will try to use this libarchive first and do some tests. Your approach may
> > > > >>collapse in case compression libraries doesn't support parallel
> > > > >>compression/decompression (I mean that you write or read several files from
> > > > >>single archive file) I would be much surprised if at least writing will not
> > > > >>work.
> > > > >
> > > > >This is a likely issue, but fortunately it would not prevent all use cases.
> > > > >
> > > > >>I wonder if there is other solution: zip could be protocol as it is now, it
> > > > >>allows to benefit from list API and gives flexibility other demuxers to
> > > > >>benefit from it. There could also be a "directory" demuxer which would also
> > > > >>use that API and possibly could serve streams in your way. That demuxer
> > > > >>could also handle directories over any protocol that supports that API.
> > > > >
> > > > >That was the kind of idea that I had. But I believe that to get that working
> > > > >a bit reliably, we will need to extend the directory listing callbacks to
> > > > >allow a URL context to create new URL contexts, to open remote files without
> > > > >establishing a new connection (and it will also be necessary for network
> > > > >servers). Some kind of VFS API, then.
> > > 
> > > Agree.
> > > 
> > > > This can be even harder as opening archive file require stat or other smart
> > > > way to check some candidates that ought to be a archive file. See below.
> > > 
> > > > >>>ffplay zip:///tmp/outer.zip/tmp/inner.zip/tmp/data.bin
> > > > >>libzip can't handle it (the same way it cannot handle files via protocols),
> > > > >>maybe libarchive will be better
> > > > >
> > > > >I think you misunderstood the question. I was not asking whether it would be
> > > > >able to decode nested files, but how your code did split nested paths: would
> > > > >it try to find /tmp/inner.zip/tmp/data.bin inside /tmp/outer.zip, or
> > > > >/tmp/data.bin inside /tmp/outer.zip/tmp/inner.zip (assuming someone was
> > > > >stupid enough to name a directory dot-zip)?
> > > > 
> > > > I assumed it is local file (no other option so far). So I stat full path
> > > > (/tmp/outer.zip/tmp/inner.zip/tmp/data.bin) for being a file, if so then I
> > > > opened it as zip file and used fallback to open first file.
> > > > If not then I stat by path components: /tmp/, /tmp/outer.zip, and so on...
> > > > /tmp/outer.zip is a file so I open it and treat rest of the uri as a path
> > > > inside zip.
> > > 
> > > walking the path means that the archive protocol must know about the
> > > syntax of the underlying protocol (file, http, ftp, etc.). that seems messy.
> > > also inefficient if you have got to walk a long ftp path.
> > > 
> > > wouldn't we be better off defining a special character that seperates the zip
> > > path from the inner path. obviously we'd need some way of escaping the character
> > > if it is legitimately part of either path.
> > > 
> > > ffplay /tmp/amovie.zip
> > > ffplay http://subtitles.org/amovie.zip#amovie.srt
> > > 
> > > the syntax should support nested archives (even if the libzip/archive cannot handle
> > > that yet). e.g.
> > > 
> > > ffplay /tmp/amovie.rar#/tmp/amovie.zip#amovie.srt
> > > 
> > > -- Peter
> > > (A907 E02F A6E5 0CD2 34CD 20D2 6760 79C5 AC40 DD6B)
> > 
> > No. '#' is perfectly allowed in URLs and local filenames.
> 
> of course it is, hence my qualification above:
> >> obviously we'd need some way of escaping the character
> 
> so if '##' reduces to '#', then:
> ffplay /tmp/amovie##1.zip#amovie##1.srt  would open 'amovie#1.srt' inside '/tmp/amovie#1.zip'
> 
> '#' was only given as an example.
> pick a character (or character sequence) that is easy to type, but infrequently used in
> uris/filenames, such to avoid having to escape to often.
> 
> -- Peter
> (A907 E02F A6E5 0CD2 34CD 20D2 6760 79C5 AC40 DD6B)

There's no such character - you can _always_ clash with unix filenames.
The only way to disambiguate this is using a protocol prefix.