[Haskell-cafe] Uniting graphics libraries

[Tillmann Vogt]

  Am 18.09.2010 15:14, schrieb Vo Minh Thu:
> Hi,
> This is a great goal! I've also been thinking in solidifying all
> things 3D on hackage lately and forming a game and graphics strike
> team. (The idea is that even if you're not interested in games, there
> are still a lot of common things.)
>
> Now it seems you see Collada as a good common starting point. I don't
> know much about Collada and I can't really say if it is a good idea or
> not. But by following the blender development mailing list, it seems
> people don't really 'trust' it, in the sense that in practice, it is
> still difficult to move things around between different programs
> through Collada...

Maybe the format isn't properly implemented in some programs. I have 
also experienced examples-files that didn't load.
But it think that collada is currently the best format and it's quite 
powerful (shaders, physics, ...).

> Also Collada is (I am not sure) just an interchange format and you
> talk about data types. Can you be a bit more specific about what you
> envision? Are they a direct representation of Collada?

Yes, pretty much a direct representation. But some things can be made 
simpler. I.e. I replace the <instance_> tags with their value and have a 
type garantee instead of trusting that a url referenced object exists.
The Collada people stress the point that its only an interchange format.
But Google is using it also for delivery (Google Earth).
If one can live with some seconds longer loading it is no problem.
By the way wonder why Collada isn't advocating binary XML?
That would make things faster.

> You talk about combining the different libraries on Hackage, would you
> like to do it through Collada?
Yes, the types . I currently don't see a better way.
> When I said I thought about solidifying things lately, I was thinking
> to the problem you describe but at a lower level: for instance there
> are many different representations for 3D vectors and transforms. Is
> it also a concern for you?
This is a problem. But I would accept a majority vote. At the moment I 
would use the same vector library as gpipe.
> Cheers,
> Thu
>
>