[matroska-general] Re: OGG Vorbis ( ;-) )

Christian HJ Wiesner chris at matroska.org
Tue Feb 25 13:44:26 CET 2003

Steve Lhomme wrote:
> Early tests with MatroskaDub (not mine) show that video+vorbis is smaller in
> matroska than in ODD. It would be interresting to test wether it's also the case
> with Vorbis only. If so, that will be a good reason to push matroska in the
> hardware world as soon as possible, since ODD Vorbis is getting hardware support
> these days, I'm sure users would like to save a few more bytes (octets ;) on
> their players.
> http://news.zdnet.co.uk/story/0,,t269-s2131016,00.html?rtag=zdnetukhompage
> http://www.matroska.org

I am sorry Steve, but i dont agree here. We should have the highest 
respect of Monty's work on Vorbis, and if he as the main developer 
decided to make his own container for Vorbis ( dont forget, matroska was 
not available at that time ) thats ok with me.

Now, you will say whats wrong about offering the users the choice in 
what container they want to put their Vorbis streams, and i do agree 
with this point of view.

Having said that, the downside of doing so is that currently the old, 
outdated MP3 standard is still being used by most people, and not Vorbis 
! There are a couple of reasons for that, mainly
- people dont know about Ogg Vorbis
- people dont care, MP3 is fine for them
and, most important :
- !! Ogg Vorbis is not supported in most applications !!

Now, if we advertised to put Vorbis audio streams in matroska files 
instead of Ogg's, this would lead to

- confusion. Normal people dont know what a container is, and they arent 
interested either. As a result they would decide to go on using good old 
MP3, as its compression format and container in one and the same thing, 
as easy as it can get. Nobody will care about a few bytes less overhead.

- lack of interest from the hardware companies to add support for 
Vorbis. They would think that the opensource community is much to 
frequently changing, and fear the risk to fail with their development 
costs, so they prefer to support AAC/MPEG4 instead. This cant be our 
goal, not at all.

- anger from the Xiph people against our project. True, Emmett didnt 
treat us correctly, but he is just one person of a big, successful 
project, and these people are very dedicated and created an excellent 
audio codec, at least for lower bitrates. IMHO it would be a complete 
disrespect of their work if we critized their container publically and 
were trying to convince people to use matroska instead. UCI as an 
interface in standard vorbis.dll would stay nothing more than a dream then.

Now, video is a completely different subject of course !! I will fight 
hard to get hardware support for matroska, and i already tried to make 
contact with 2 companies ( KiSS and Neuston ) seeling DivX/MPEG4 capable 
units, but they werent interested in our current state of development at 
that time. Rest assured i will contact them again once we have working 
software to present ;).

I can see a few important aspects why hardware vendors could be very 
much interested to implement matroska in their standalones :

1. AVI sux. There are too many different applications creating 
non-spec-compliant AVI files, and there are simply too many hacks 
floating around, so its very unlikely those standalone units can play 
*ALL* AVI files fine. Sync is more than questionairy, although all of 
them claim they can support VBR MP3 also, and even multiple audios 
treams ( i doubt it ).

2. Subtitles : There is no subs specs for AVI, instead there are many 
different, completely incompatible subtitles formats floating around, 
sometimes even in the AVI file ( complete hack, breaking specs ) or as 
outside files with same file names and different extensions ... a pain 
in the a** to support for a hardware device.

3. DivX3 / Mode2 Form2 : All existing hardware players are based on the 
SIGMA EM 8500 decoder chip now, and this is a MPEG4 ISO implementation 
and cant deal with M$ MPEG4 V3, the basis for DivX3, so all DivX 3 
movies are unplayable.  -h  from the XviD team is already working on a 
converter to transcode MPEG4 V3 into MPEG4 ISO in a LOSSLESS process ( 
yes, this is possible, main code taken from ffmpeg.sf.net ), without 
reencoding, and we know that DivX Networks is doing the very same thing 
right now. Unfortunately, he mentioned that the files may be 5 - 15% 
bigger in the end, as there are some major differences between M$ MPEG4 
and ISO, and this will lead to bigger file sizes.
What seems to be a big disadvantage could turn out to be a big win-win 
situation for matroska, as mode 2 form 2 burning can offer 100 MB more 
on a normal CD ;) !!
In the end, i could see a situation where the hardware manufacturers 
will provide an DivX3/AVI to MPEG4/matroska converter tool with their 
players, so people can convert their DivX3 movies into something 
playable. Note that the service teams of these companies must have a 
hard time currently, trying to explain people why DivX3 is NOT DivX :-D !!

4. Missing Vorbis support : Vorbis in AVI is a no-no, the future of OGM 
is at least questionable ( also there are no docs for it ) and there is 
still no official announcement from Xiph people that Ogg Theora will 
support MPEG4 video and AC3 audio, being a must for the companies 
offering standalone units. So matroska is the obvious choice !!

I hope you dont mind me critizing you on a public ML, but as i know you 
are an open person and for sure you would do the same with me also :P




More information about the Matroska-general mailing list