[vorbis-dev] Re: [vorbis] Finally, it can be told!

[Ignacio Vazquez-Abrams]

On Fri, 31 Aug 2001, Jack Moffitt wrote:

> Any game that is this critically speed dependent, would probably not use
> python in this way, unless python has suddently gotten much faster.
>
> I'd love to use python as my only language, but speed is where these
> things break down.  I can certainly see managing game pieces in python,
> but I think games will always have some compiled language as teh base,
> which they expose to the pythons, tcls and perls of the world.
>
> I can vouch for the fact that vorbis-python is quite fast.  It's not
> even significantly different than ogg123 itself.

Certainly while Python has gotten faster from 1.5.2 -> 2.1.1, it is not as
fast as C.

> Also, there is more to consider.
>
> there is also an ao-python and an ogg-python, neither of which you have
> a counterpart for.  Python's standard library doesn't do sound output
> particularly elegantly, and I think ao-python or libao itself does a
> much better job.  I can understand wanting a more python friendly
> interface for libao, and that's a reasonable goal for ao-python.

"Doesn't do [it] particularly elegantly"? You're being kind. Python is
terrible for that sort of thing.

> ogg-python, while not strictly necessary just for Vorbis decode, _is_
> necessary for other mediatype and will be necessary for structured
> metadata.  To assume that vorbis the only thing here is incorrect.  We
> are building up a nice cohesive frameware into which vorbis fits.
>
> Currently ogg-python is little more than a placeholder, which is fine.

I understand the larger picture of Ogg and how Vorbis and Tarkin relate to it.
What I have been creating works with what we have today. I certainly don't
mind extending it if that happens to become necessary. However (and this is
BIG...), if you think Python is bad at audio output, you should see it do
video output!

> Both of your modules need sound output, and I think that sound output
> should be ao-python.  If ao-python is not python-friendly enough, I see
> no reason why we can't all agree on an API that _is_ python friendly.

If you take a look at testvorbis.py included in my package, you'll notice that
it does in fact use ao-python for audio output. I definitely agree that we do
need some sort of API for Python audio input/output. However, we should
probably develop it in conjunction with Van Rossum et al.

> Vorbis _should_ be layered on the sound output module, since there are
> many more things you can do with it than just output sound.  To reduce
> vorbis' utility to a small domain of problems seems like a bad idea.

Mmm. I think you misunderstand my meaning of "layering". From Andrew Chatham
<andrew.chatham at duke.edu>:

"It probably wouldn't be very hard to implement most of this in actual Python
script on top of the existing modules..."

The sense of layering that I'm using is in creating an Adapter class from the
low-level vorbis-python to the wave/sunau API.

> Granted, I haven't dug deeply into the source of either module, and I
> haven't used vorbismodule at all, but I do think we should try to work
> together, if only to keep from confusing the python community and making
> more support work for our core team (hell, we already answer plenty of
> question about rio's, winamp, and other people's products).

Agreed. The less work we have to do all around the better.

Something that might be looked at is starting up a Python Multimedia SIG
(http://www.python.org/sigs/). That way we could develop consistent APIs for
both audio and video.

> So, is ao-python not suitable for your purposes?  It seems like it would
> be a great replacement (assuming we put in win32 support) for all the
> disparate audio modules python currently employs.

Absolutely. I just think we need to work through API problems and extend
compatibility beyond Linux/*BSD/Win32 to the commercial Unices as well.

> jack.

I actually only started work on vorbismodule on Sunday night, so I certainly
see room for improvement. Hopefully we can all improve everything together.


-- 
Ignacio Vazquez-Abrams  <ignacio at openservices.net>

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-dev-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.