ogg dev - Feb 2008 - Ogg bitwise.c bit tracking

On 09/02/2008, Ralph Giles <giles@xiph.org> wrote:
> Yes, all that is reasonable. The original argument was that we need a
> library anyway, no one noticed the function call overhead, and using
> libogg simplified the embedding.
>
> Now, we'd like to remove libogg entirely as a dependency to avoid
> confusion when using other containers, but that's an api change...
As far as libvorbis is concerned, wouldn't the step of copying the
bitpacker into libvorbis, where the functions are only used
internally, just allow a dependency change?

There is of course the bit of the libvorbis API which talks in terms
of ogg_packet structures, but maybe we could replicate the definition
of that in a vorbis header file (inside of #ifndef _OGG_H).

Removing the bitpackers from libogg would of course be an API change
to libogg, but we could defer that.

K.

On 2/8/08, Conrad Parker <conrad@metadecks.org> wrote:
> There is of course the bit of the libvorbis API which talks in terms
> of ogg_packet structures, but maybe we could replicate the definition
> of that in a vorbis header file (inside of #ifndef _OGG_H).
If you think that's safe, then let's do that. (assuming monty
doesn't yell.)

 -r

[08:41] <xiphmont_> If there was a clear problem with incompatable
non-Ogg implementations we could reasonably help resolve... But I
think the cows are already crushed under the barn doors, or however
that saying goes.
[08:41] <xiphmont_> Eg, I think it is a slightly higher priority to
make is more clear how to use the codec libs without libogg.
[08:42] <xiphmont_> That is, the bitpacker is built into libogg
although it really doesn't need to be.
[08:42] <xiphmont_> Still it is, and that gives the impression our
codecs require Ogg.
[08:42] <_Ivo> how do you suggest this to be done?
[08:42] <xiphmont_> when really what they require is a bitpacker
according to the spec.
[08:43] <xiphmont_> JM and I discussed it on the Quebec trip
[08:43] <xiphmont_> I think to make it clear we split libogg into
libogg and libbitpack or something
[08:43] <xiphmont_> and also rejigger the spec to seperate out the
bitpacker specification into another document.
[08:43] <xiphmont_> No *actual* changes to code flow.
[08:44] <xiphmont_> Just how the namespace and information is presented.
[08:44] <xiphmont_> At that point, -logg wouldn't be needed to build
eg Vorbis and Theora
[08:45] <xiphmont_> In short, an exercise in clearing up a misperception.
[08:45] <_Ivo> so, in theory, it's something feasable and that can be
done quickly?
[08:46] <xiphmont_> Anyway, I'm just bringing that up as something
that is being handled.
[08:46] <xiphmont_> it is on the development queue, not particularly
high priority.  But in itself, a simple thing, yes.
>From [1].  Read libbitpack as libvorbis and we have a happy ending.
[1] http://xiph.org/minutes/2007/xiphmeet_20070711.txt