Re: [openvrml-develop] OpenAL sound support in OpenVRML

[Braden M. <br...@en...>]

On Mon, 2002-04-08 at 10:41, -oli- wrote:
> As I wrote some weeks ago, I'm going to integrate OpenAL-support for the 
> sound in OpenVRML.

Okay. I'm gonna be a bit more explicit than I was before in order to
ensure expectations are set appropriately. I don't want you to complete
this project and then be surprised by its reception.

I don't think OpenAL is the best solution for audio in OpenVRML. At one
time it looked like it might be; however...

        * OpenAL's primary user and developer (Loki) is no more.
        * There was never a (successful?) attempt to package OpenAL in a
          way that would be palatable to end-users and OS distributions.
          Even now, the OpenAL Web site directs persons looking for
          "1.0" to CVS. The project has reached 1.0, and there's *still*
          not a tarball.

So OpenAL is difficult for end-users to install, and prospects for
continued support and development seem dim.

What this means:

        * *Yes*, I would accept a patch for OpenAL-based audio support.
        * This support would be disabled by default. If someone wanted
          to build with it, they'd need to configure with
          "--with-openal" or somesuch.
        * If OpenVRML also had support for some other audio solution (I
          mentioned GStreamer before, and I still think it's our best
          bet under Linux), I would not be inclined to spend too much of
          *my own* energy maintaining multiple audio solutions in
          parallel.

If you think I'm trying to steer you toward GStreamer instead of OpenAL,
you'd be right. However, I appreciate that you may have your own
motivation for wanting to use OpenAL, and I have no intention of
belaboring the issue beyond this posting. I appreciate that your work
*will* improve OpenVRML, even if it's not exactly what I'd do. :-)

> Here I have some questions regarding this:
> 
> (1) In which method(s) can I implement the assignment of the spatial 
> parameters (position, direction) of the sound to the OpenAL properties? 
> NodeSound::render() or NodeAudioClip::update() or NodeAudiClip::render(), 
> which doesn't exist yet.

That logic belongs in the Sound node.

> Regarding this, the method should be called 
> whenever the position of the Sound-Node or the ViewPoint-Node changes. 
> Besides that, also the time-dependend functionality in NodeSound::update() 
> should be maintained.

In the old architecture, that logic would belong in VrmlScene. In the
new architecture, it belongs in SoundClass.

> (2) The WAV-File is read totally into memory at the moment (see audio.h) 
> and then put chunk by chunk to the dsp-device (see NodeSound::update()). Is 
> this right?

As long as we aren't bothering with streaming, I think that's
acceptable. However, I'm not much of a "sound guy", so I might not know
what I'm talking about here.

-- 
Braden McDaniel                           e-mail: <br...@en...>
<http://endoframe.com>                    Jabber: <br...@ja...>