Subject: Re: [linux-audio-dev] Re: New OSX services & LAAGA
From: Francois Dechelle (Francois.Dechelle_AT_ircam.fr)
Date: Wed Jun 06 2001 - 15:00:06 EEST
Paul Davis wrote:
>
> Thomas Hudson wrote:
>
> >What I thought was especially interesting is their AudioUnits API. Seems
> >like they're specify a unit generator or LADSPA-type plugin system at the
> >OS level, though it is a pull I/O model.
>
> I hope everyone took the oppurtunity to review Apple's specs. I have
> not, because I don't want to be "contaminated" at this point. Because...
>
> I am *VERY* excited.
>
> Two days ago, I sat in the Piazza della Republica in Firenze with
> Francois Dechelle (IRCAM/jMax), and we talked in great detail about
> the whole LAAGA issue, plugin/component models and the OS-X audio API
> (Francois has already had extensive experience (relatively speaking)
> with the OS-X audio API).
>
> I believe (and I think Francois does too, but I don't speak for him)
> that there is now ample justification for every element of LAAGA's
> conceptual design to become the primary audio API for Linux. The
> callback model (Thomas calls it a "pull model" above, but I prefer
> "callback" since for capture this is actually a "push model") is
> firmly entrenched in ASIO, every plugin system under the sun including
> DirectX, and now OS-X. It may even be part of the BeOS audio API. The
> use of floats as a universal standard for audio data is also present
> in most other APIs, including IRIX which for years was (and perhaps
> is) by far the most advanced Unix-like OS when it came to
> multimedia. Finally, the model that most applications (generally) do
> not alter hardware settings such as buffer size, sample rate and so
> forth, is well established on all versions of Windows and MacOS (*)
>
> During the somewhat arduous 2 day journey back from Firenze, I have
> written most of the code to define and implement an extremely simple
> audio API that embodies most of what we've spoken of here over the
> last few weeks. During the day today, I hope to type it in and send it
> in for your review.
>
> The work that Abramo has done on the ALSA PCM "shm" device will ensure
> that applications which do not wish to adopt this model will be able
> to work even on a system that has a LAAGA-like audio engine
> controlling the only audio interface. If it was not for his work on
> this, it would be much harder to propose the fundamental change in
> audio APIs that I'm about to do.
>
> I'd also like to note that in the last 4 weeks I have actually managed
> to have face-to-face meetings with:
>
> Dave Phillips
> Richard Furse
> Francois Dechelle
> Maurizio Umberto Puxeddu
> Gunter Geiger
>
> and others. The sense that there really *is* a Linux Audio Developers "group"
> feels much more tangible to me right now.
>
> excitedly,
> --p
>
> (*) I actually think that ALSA may have made a basic design error in
> this area, but its not a serious one in that it can be conveniently
> glossed over in various ways. ALSA needs to be a general purpose API
> anyway, and in that sense its probably better that it continues to
> support the OSS tradition of "open; set parameters; use; close" rather
> than enforce the "open; use; close" paradigm.
Hello,
Yes, it was a very interesting discussion.
I've already used part of the Apple API, namely the HAL (Hardware Abstraction
Layer) for a first rought porting of jMax to Mac OS X. It is based on callbacks,
the callback being called in a separate thread (this is not documented at all).
There is something that is not clear from what I read: is the "AudioUnit" API
for inter-applications communications, or is it a plug-in API ? It looks
more like a plug-in API.
The most interesting thing, IMHO, in what has being discussed yet about LAAGA,
is the idea of inter-applications (i.e. inter-process) audio interchange. As far
as I know, the only system on which this has already existed is BeOS.
Does anybody know more about it ?
François
This archive was generated by hypermail 2b28 : Wed Jun 06 2001 - 16:43:27 EEST