[Bf-gamedev] Proposal to (try to) get better FBX support (Nahuel Belich)

[Bastien Montagne]

I do not think Collada is the way to go, I would even say it’s the other 
direction we should take, if we were to work on a new format.

I would go to a minimalistic format. FBX is much better than collada on 
this regard, but it is cluttered with an history of crapiness - and some 
aspects (like its object transform model) are still crazy.

An exchange format does not have to support every possible feature. An 
exchange format has to be simple!

E.g. I like Wavefront .obj format - it is simple (even though Max 
managed to break it with odd expectations on smoothgroups :/ )!

So if I where to design a new format, I would take the basics of FBX, 
and get rid of everything I do not like in it (the multi ways to 
represent an armature and skinning, the insane object transform 
handling, complex anim curves, etc.) - not to mention stupid unused 
things like custom shaders, etc.

We should even probably build it bottom-up: we start with a *very* 
simple base, e.g. object + (mesh) geometry, and then build upon this, 
step by step, with each time wide consultation of all interested 
entities & people to know whether this is *really* needed.

And having a system of custom props, for people who really want to write 
more specific/advanced data.

Just my two cents, imho time needed to design and get accepted a new 
format would be really huge, this won’t work unless a good part of the 
CG/Game  community (including some big companies) step in - and I have 
the feeling most are more or less satisfied with current status… :/

Bastien

Le 17/09/2014 09:56, Jens Christian Restemeier a écrit :
>
>     I had tweet this question, is it an insane idea? to contact the
>     the mayor game engine companies (udk, unity, crytek, tell me
>     another important) and others 3d apps to, and start some sort of
>     new open exchange format?.
>     Autodesk put the fbx in the market because need to expand
>     conection among "their" aplication and even among those, took
>     years to take info correctly from one application to another, and
>     the rest took that because it was there, im sure that many main 3d
>     sofware companies have similar problems trying to deal with the
>     close fbx sdk.
>     This where my two cents, for the moment im using blender fbx to
>     export to unity but at the same time i dont like to see effort
>     waste in something that may change and become obsolete with future
>     changes on a close code that its so hard to tackle.
>
>
> That already exists: https://collada.org/
> I generally like the idea of Collada, but it has a few major problems:
> It is hugely over-engineered: You can write any data that a modelling 
> application may want to write, and any data that a rendering engine 
> may need. While that is a good thing for an exporter, an importer has 
> to support everything and translate this into an application format, 
> or you would need intermediate tools to massage your data into a 
> format you can use.
> Applications have varying levels of support: There is no reference 
> implementation (that I know of). The official SDK is only a DOM loader 
> (last time I checked). It's not as easy to use as FBX where you plug 
> in your model file and get back a scene graph. I've got a file with a 
> few separate skinned models that works fine in Maya and Unity3D, but 
> Blender fails to find the bone-to-deformer info. (I haven't checked if 
> that is because of my exporter code or a bug in the Blender importer. 
> It fails to match SID identifiers.)
> The shader model is massive overkill: You can include everything from 
> high level blinn/lambert shaders to a low level cgfx-style rendering 
> description with passes and render states. And while I like the split 
> into materials and effects, applications that don't support this don't 
> seem to map this cleanly to their system.
> Collada supports many ways for blocks in a file to reference each 
> other other even external files. That may be hard to map into an 
> application-internal format if you want to keep them intact for an export.
> Some people I spoke with are not happy with the underlying XML format, 
> and who would prefer a binary format. I don't really care that much, 
> because it is easy to load and write XML, but I see their point if you 
> want to support huge models.
>
> I think there are a few other attempts at open interchange formats, 
> but they don't seem to have much traction so far.
>
> Of course, if you want to put in the effort I would try to polish the 
> Collada code instead of inventing another format.
>
> Jens
>
>
> _______________________________________________
> Bf-gamedev mailing list
> Bf-gamedev at blender.org
> http://lists.blender.org/mailman/listinfo/bf-gamedev

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.blender.org/pipermail/bf-gamedev/attachments/20140917/33f365ef/attachment.htm