[Bf-committers] Render API follow-up

[Ton Roosendaal]

Hi all,

For many years we try to settle the render API topic, especially to 
allow advanced export and/or integration with third party renderers. 
For various reasons it remains getting postponed; mostly because the 
specs for such a system are complex and fuzzy.

Before Wintercamp I was contacted by Terrence from Lux to at least give 
some roadmap or insight in how we like to proceed with rendering in 
Blender. Many developers out there are eagerly waiting for such 
clarity, and they are getting a bit tired or demotivated with our slow 
progress in this area. :)
I've promised we would spend some time together during the Wintercamp 
workshop on this.

We've discussed a couple of strategies; including a more 'high end' 
integration, not based on exporting geometry/materials but for example 
exposing an api to use raytrace calls and visibility info.
However, what we quickly fell back to was the simple fact that we first 
should work on a good (re)design of how Blender will tackle rendering 
in general. This should start with a better way to define own shaders 
(real shader tree editor, shader language), but also a more advanced 
interal shading/lighting pipeline, and better methods to handle caching 
(geometry bucketing, image tiling, shadowbuffer tiling).
This will, combined with the 2.5 flexibility to extend on Blenders data 
via RNA, deliver a much better starting point for a future proof design 
for integration (plugins) or export.

So... not an easy answer! :) But I think it's fair for Blender to first 
fix their own  issues before moving to get better support for external 
renderers. That topic is really high on the prioriry list, and can get 
first priority when 2.5 is in a workable state, when we start up our 
next movie project for example.

It's extremely tricky to pin me down on dates for this, but very likely 
work on rendering code comes back before july, with a working new 
architecture for this some months later (october?).
Designs for an improved render system could start away though, and I'd 
be very interested to see what Terrence or others have in mind, 
especially how integration should ideally work from their perspective.

-Ton-

------------------------------------------------------------------------
Ton Roosendaal  Blender Foundation   ton at blender.org    www.blender.org
Blender Institute BV  Entrepotdok 57A  1018AD Amsterdam The Netherlands