[Openal-devel] New AL 1.1 Spec and Beta Windows Implementation

[Alexandre Mah]

On 30/04/2005, at 6:40 AM, James Tomaschke wrote:

>> I think it would be nicer for the default to be 1.0 and for 
>> applications that want to use doppler with a velocity unit other than 
>> the mach to set the parameter explicitly, instead of having an ugly 
>> constant like 343.3 being introduced into the specification as a 
>> default.
>
> Basically I totally agree with your statements, but I think that since 
> the spec authors made a compromise and removed 343.3 from a constant 
> to a parameter (because of some of the issues that we brought up),  we 
> could be a bit flexible as well and pick our battles.

Well, we do have the OpenAL 1.0 specification on our side, which I 
think counts for a lot.

I can understand that the Windows implementors don't want to break 
compatibility with their (incorrect) Windows OpenAL 1.0 implementation.

So here is one possible compromise.
How about we specify the default value for AL_SPEED_OF_SOUND as 
"undefined" rather than 343.3?

This means that:

1.  Windows, Mac OS X, and Linux OpenAL 1.1 implementations have the 
ability to retain compatibility with their AL 1.0 implementations.

2.  Any programmer using Doppler in OpenAL 1.1 will want to explicitly 
set AL_SPEED_OF_SOUND to whatever value they find meaningful.

3.  This will still allow a future version of OpenAL to specify an 
explicit default value for AL_SPEED_OF_SOUND, at a future point in time 
when no OpenAL programmer is relying on any default value for 
AL_SPEED_OF_SOUND (since the default is specified as "undefined").

4.  There is no ugly constant 343.3 in the specification.  (If we put 
it in the specification now, it will be there for all of eternity and 
we will never be able to get rid of the hideous wart.)

How does this sound to people?

I think if we do fix a constant for AL_SPEED_OF_SOUND rather than 
specify it as "undefined", it should agree with the OpenAL 1.0 value of 
1.0 rather than the Windows implementation value of 343.3, since I 
really don't think conformant implementations should have to suffer 
because of non-conformant ones.