Arc Riley wrote:
> I've decided to rework file loading based on the difficulties I've had.
>
> First, libogg2 works great, but upgrading libvorbis (etc) to use it is
> outside
> the scope of this project and just adds more to my/our plate pre-1.0 release.
>
> Of course we should upgrade to libogg2 the moment these codecs are available
> using it.
>
Yes, and using two versions of ogg at the same time might be problematic.
> Second, we have numerous problems embedding Soy data in Ogg chains. We need
> to know, within the Python __new__ function for File/etc, what classes are
> being used so that we can initiate the instances of them. That means that
> __new__ would have to load the entire file at once, potentially interupting
> the render cycle and killing the benefit we'd otherwise gain from progressive
> loading. Coreloop can ADD data to any class already created but cannot
> create
> new instances itself.
>
> Yes - we could say "no chaining" - and setup other rules, but it's gotten
> silly. Most of all, in using Ogg for storing this, we loose the ability for
> Blender (and other 3d editors) to load/save the data without PySoy. While
> having PySoy around can greatly improve it's speed it shouldn't be a
> requirement.
>
True.
> That is - if someone wanted to load a soy.bodies.Mesh from a .soy file to use
> in their CrystalSpace-based (GPLv3 licensed) game they shouldn't have to
> install PySoy just to read this data into Blender.
We'll miss you (a little), ogg2. Good choice, arc.
_______________________________________________
PySoy-Dev mailing list
[email protected]
http://www.pysoy.org/mailman/listinfo/pysoy-dev
