[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Fwd: Re: [AVT] Re: New I-Ds for Ogg technologies (Vorbis overRTP, Ogg file format, Mimetypes)





--On fredag, januar 24, 2003 06:58:38 +0000 Michael Smith <msmith@labyrinth.net.au> wrote:

  It's pretty clear that OGG players can identify which codec is being
  used. But this document does not say which bytes of the BOS page can
  be used to identify the codec. (first 4 bytes? first 16 bytes? up to
  the first NUL?)

Again, based on vorbis as an example (for the spec doesn't explicitly
state this), any leading part of the data (i.e. packet contents, so these
are not the absolute leading bytes in the stream, since there is the page
header first) in this first page. Vorbis uses the first 7 bytes (0x01,
0x76, 0x6f, 0x72, 0x62, 0x69, 0x73 -- which is a leading 0x01 followed by
the ascii values for "vorbis"), but the intent is to not have a single
fixed-size identifier. We should, however, specify a _maximum_ length for
this).
followup..... the IETF's been dealing with multilevel protocols for some years, so identifying the "protocol at the next layer" is something we're used to... not trying to say what's the right way, just musing about the problem....

you've basically got two problems:

1) how to make sure there's information enough to identify the codec, and information enough to know when the codec has been uniquely identified?
(it would be silly if players refused to play 0x01 + "vorbis" + 0x01 because all the sound they had been tested on had a 0x00 in the position after the "vorbis" string, and the player checked the values of 8 bytes, for instance; otoh, 0x01 + "vorbiz" is probably a different codec, so you should check at least 7 bytes, not 6)
(btw, my Linux Mandrake "magic" file claims that the byte after "vorbis" is the 32-bit version number, and zero is the only known value, I think...)

2) how to make sure no two codecs pick the same identifier string?
(the "registry" problem)

the last is usually not a problem until some name becomes "popular" - but think about the number of formats that have been using the .doc file name ending on the Windows platform for an example of collisions.... if you solve that problem by establishing some kind of registry (rather than "trust to luck and goodwill"), you also have a convenient place to store pointers to the specifications of those formats.....

(In my personal .ogg collection, there are no examples of anything but the vorbis codec, unfortunately... and all of it has been encoded with a single codec, so this is not much of a test suite....)

Harald


_______________________________________________
Audio/Video Transport Working Group
avt@ietf.org
https://www1.ietf.org/mailman/listinfo/avt