[ogg-dev] Proposal: An extension to rules all others

Ivo Emanuel Gonçalves justivo at gmail.com
Mon Jan 15 11:24:12 PST 2007

Proposal: An extension to rules all others

Copyright Notice
None.  Public Domain.

Extension flamewars surrounding media formats inside Ogg containers
are a pain in the behind, cause problems in the adoption of said
formats, and confuse users.  This modest proposal directed at the
maintainers of Vorbis, Theora, Speex and FLAC will suggest a solution
to solve these issues (and others).

1. Introduction
According to Lore, there in ancient days was Ogg, a container for all
kinds of media formats, but for years to come, Ogg had only one
companion, a music box known as Vorbis.  Together, the two tried to
battle the evil kingdom of MP3, and though not successful, songs were
composed of Ogg and Vorbis feats.  However, misunderstanding arouse,
and the two became associated as one entity: the OGG.

Now, fast-forward to the present.  The "OGG" misconception states that
any file with the .ogg extension is "OGG", which is some kind of
nerds-only format for music, or an MP3-wannabe.

Now, we of Xiph know this is bullshit, but this kind of mentality has
spread so widely that it's hurting the Ogg project as a whole, and its
family of formats.

2. The Problems Caused by the OGG disease
When one speaks with the average computer user, the one that may
occasionally check Slashdot (or God forbid, Digg), s/he will tell you
in the world of Open Source (or FOSS) there are two formats for audio:
OGG and FLAC.  More or less, s/he has a basic idea of what OGG is:
it's some kind of audio format like MP3, whose name sounds like
caveman speech, and nobody uses.  S/he probably likes FLAC, though.
The FLAC fad is yet to come, but its day seems to draw near.

There isn't one problem here; there's several.  The person described
above may be a blogger, but he has no idea what Speex is.  The person
above may have an iPod, and s/he won't care much for any audio format
other than MP3.  S/he won't bother to look for a hardware player with
support for other audio formats, and more importantly, s/he won't even
bother consider nagging Apple about adding support for other formats.
If this person has ever gone to Wikipedia, s/he may have come across
media files with an .ogg extension, but s/he has no idea why it's
labeled as video.  "Isn't OGG supposed to be that nerds-only music
format?  What's this all about?"

This person will try to make the video work, following the useful
instructions on Wikipedia Media Help page to install illi's codecs,
and so on.  The problem is when s/he tries to play it on any MS
Windows player that isn't VLC or RealPlayer.  S/he'll be royaly
screwed.  What with the majority of the few programs that recognize
the .ogg extension assuming it's Vorbis.

At this point, the person will simply give up.  The video wasn't that
important to being with.

However, the person may think, "Hey, other people will have the same
problem too", so s/he will go ahead and complain to Wikipedia about
it.  Now make that several people a week, and not only in Wikipedia,
but anywhere else hosting Theora videos.  And remember that it's only
the peopler that care.  Like 1% of the Internet population, maybe
less.  Nobody else cares.

I think there's a lack of consistency and correctness in how the
different developers in Xiph are doing things.  We need some unity if
we are ever to take on the world.  Not that I'm saying I'm the
Messiah, and I have the answer for everything, or that I know of all
the problems there are (or not).  But on this issue, however, I hope
to have your full attention, gentlemen.

3. One extension to rule them all
This title is misleading, because it's not possible to have one single
extension to fix the problems described above.  Especially, if one
considers this proposal aims for backward AND forward compatibility.

The proposal isn't perfect, but in my opinion it's a fine step towards
solving some of the problems plaging the projects surrounding Ogg.

I propose that each project (Vorbis, Speex, etc.) states in its
specification two file extensions, and those two extensions are to be
the only ones allowed for that given format.  Implementators are
supposed to support both extensions.  Content creators are supposed to
choose one of them according to whatever criteria they care about.

Are there arguments against this?  Of course there are.

One might say that .avi is always .avi even though it might be MPEG 1,
XviD, DivX, or anything else.  Yet, you forget that .avi is ALWAYS for
video, so most software won't complain when they see an .avi file
(except for WMP that complains for everything).  The worst that may
happen is for the software to ask the user for the appropriate codecs.
 This doesn't happen with .ogg, though, because the Ogg container
might be used for so many different purposes that programs don't have
an uniform way of dealing with .ogg.

Matroska, for instance, specifies .mkv for video/video+audio, and .mka
for audio.  This may not be the greatest approach ever, but it works
in real life, and that's what we should aim for.

And yet, there's further opposition?  Some may say "extensions don't
matter", but they forget this is a computer world ruled by MS Windows,
where the average user is not smart enough to know what's wrong when
the video or audio file s/he downloaded won't work.  Extensions are
needed.  Anyone ever saw the UNIX "file" command trying to guess what
type a file is?  Extensions are needed.

Now, assuming you guys dig what I say, there's another problem.
Choose a new extension for your project, say .vorbis for Vorbis and we
all know what happens.  Existing software will stop playing older
Vorbis streams using the .ogg extension.  The solution as I mentioned
above is to allow two extensions, no more, nor less.

Example: .ogg as default, .vorbis as secondary.

OK, but that's for Vorbis.  Why should FLAC or Speex care?  Because of
consistency.  If we want to move Ogg away from the misunderstanding
that it's Vorbis only (the "OGG"), we have to allow .ogg to be used on
FLAC and Speex files.  And because of correctness.  If we want unity
between our projects, we have to allow .ogg to be used on FLAC and
Speex files.

So, this is what I suggest:
Vorbis: .ogg & .vorbis
Theora: .ogg & .theora (or .video as I've seen suggested several times)
FLAC: .ogg & .flac (yes, I'm fully aware that FLAC can exist without
Ogg, but it's about unity)
Speex: .ogg & .spx
OggPCM: .ogg & .oggpcm
OggUVS: .ogg & .ogguvs

For mixed content of other kinds, like Writ, or crazy combinations of
the above, or external codecs (MPEG 4, whatever), one has accept the
default .ogg only, because it's not possible to state an extension for
every possible use of Ogg.  The idea is to allow dual extension on the
most important projects under Xiph.

Notice that most of the suggested extensions are outside 8.3
limitations.  It's 2007; we do not need to worry about this anymore.

That list is a suggestion, and this you are reading is a proposal.
It's up for each maintainer to decide if it's viable, or not, but I
ask you to consider it.

This proposal isn't only philosophical.  It's about making life easier
for end-users and to tell software that if it supports Speex, then it
needs to look in the current directory for both .ogg and .spx files as
source for Speex streams.

Extensions are meaningless, and .avi sure as hell isn't just "Video
for Windows", so let's make it clear what extensions are allowed for
the different projects under Xiph, even if it's crazy talk.

4. More Solutions
Ralph Giles has been studying the possibility of adding a
"disposition-type" field on the MIME application/ogg.  This is an
approach taken by the RSS project, and I believe it's another step in
the right direction.  But it's a completely different debate, though
I'm mentioning it, because this and the solution described in this
proposal should both be used.  Operating Systems don't care about MIME
types, but care about extensions.  On the other hand, HTTP servers
don't care as much about extensions, but care a lot about MIME types.

The RSS project is considering using disposition-type with Dublin Core
attributes like "moving-pictures", "sound" and "mixed".  This syntax
may not be most appropriate for Ogg streams, but the idea behind it is
good.  Let's take it in consideration.

Example: application/ogg; disposition-type=sound

One more solution to the problem of consistency may be the use of the
name "Ogg family" when referring to the media formats that use Ogg
containers.  I think it will help fortify the ties between the
different projects under Xiph, and make easier my job, which is to
promote those projects.  Ogg family.

5. In Conclusion
Hi, my name is Ivo Emanuel Gonçalves.  I'm new to Xiph.Org, and mostly
unknown.  My project is called SpreadOgg.  It's an attempt at doing
whatever it's needed to promote the media formats under Xiph.Org.

It was hard for me to write this proposal.  Very hard.  Because it
forces change, and change is never welcomed.  Because it's bold, and
boldness tends to attract enmity.  Because it's probably going to be
ignored.  And because it may add more complexity to an issue that is
already complex.

But in the end, I thought this issue was important, and here we are.
From what I've heard, whole flame wars regarding extensions have
happened before, and they led to nothing.  Why not end this issue once
and for all, and go on with more important matters?

This message is public and is made available in ogg-dev, vorbis-dev,
flac-dev, speex-dev, and theora-dev mailing lists.  Although it's
directed at the maintainers of these projects, casual readers of the
mailing lists may discuss this proposal if they wish, but please make
sure the contribution is significant and/or useful.  More ranting is
pretty much unnecessary.

Que seja o que Deus quiser.

More information about the ogg-dev mailing list