Leider wird die nächste bzw. spätestens die übernächste Rechner-/Grakageneration nix mehr mit DX7 anfangen können, insofern wirst du irgendwann nimmer drum herum kommen.

Das waren noch Zeiten, als DX komplett Abwärtskompatibel war


Naja, leider ist an dem was du sagst wahrscheinlich wirklich was dran... obwohl Microsoft damit imho einen seiner Haupttrümpfe verspielt: Abwärtskompatibilität ohne zusätzliche zwischenschichten oder so.. na, was solls

Es hat schon seine Gründe, warum Microsoft immer mehr auf Abwärtskompatibilität verzichtet.
Unter Anderem -> DLL Hell

Hi,

Ok, to clarify things a bit:

* Priority 1 is writing a new 3d lib, with an underlying 'b3d-like' design.

* Priority 2 is wrapping it up as a Max module.

* Priority 3 is wrapping it up for other languages.

When I mentioned 'Blitz3D2' below, I was thinking about swapping out the current 3d engine in Blitz3D for the new one - language/IDE remain the same. Haven't given this much serious thought (and probably shouldn't have mentioned it) as Max/3D kind of renders the idea redundant. Still, there may be people who prefer the simpler b3d language and ide? Community response so far seems to indicate this is indeed so, but it still a 'maybe' for now...

And...there is definitely NOT a new language coming!

Some tech stuff (all subject to change of course):

While this engine wont be identical to b3d, there is one thing I want to recapture from b3d: the 'everything is an entity' concept. To be more precise - 'everything with a matrix (ie: with position/rotation) is an entity', and as such has a single parent and 0 or more children.

This breaks many of the 'design rules' I've read over the years regarding scene graphs, but is just sooo sweet to actually use, and that's gotta count for something! I have tried various other approaches and messed with other systems, but these generally lead to a major increase in API complexity for a dubious amount of efficiency benefit.

Physics-wise, Max3d will have ODE support built-in. Support for other physics engines may or may not happen. So far, I've got body (things with mass), geom (collision shapes) and ball-joint entities happening. Currently, they rely on parent/child relationships to decide what to 'connect' to - eg: Adding a child geom to a parent body will link the geom to the body (and adding another child geom to THAT geom will too, etc) - but this could change in future if necessary with 'SetBody' style calls.

Some might find the idea of bodies, geoms and joints being actual scene-graph nodes a little unusual. But in practice, I just find it much nicer to use than a system that separates out the physics stuff - you can sort of 'construct' things in your head more easily.

Rendering wise, the surface/mesh system is also likely to change a bit. B3d surfaces actually contain a 'hidden' internal mesh (b3d 'meshes' are really a container 'meshmodel' for multiple surfaces) and I want to make this side of things public.

This allows for more flexibility painting the same mesh different ways - which is important because the old 'EntityColor/EntityTexture' stuff *will not* be happening. This basically involved 'blending' multiple brushes, but was very clumsy internally, introduced a nasty rendering bottleneck and would be practically unworkable in a multi-shader environment.

In fact, surfaces are currently 'proper' entities and exist solely to connect a mesh to a material (and provide a matrix). If you want a multi-surface model, you need to create a pivot and add a bunch of child surfaces to it.

Bottom line, code to knock up a cube+physics currently looks a bit like this:

mat=CreateMaterial()
mesh=CreateCubeMesh()
surf=CreateSurface()
body=CreateBody()
geom=CreateCubeGeom()

SetMaterial surf,mat
SetMesh surf,mesh
SetParent geom,body
SetParent surf,body

Zudem schreibt er, das es keine neue Sprache geben wird, er aber darüber nachdenkt die aktuelle B3D Engine in B3D selbst mit der neuen Engine auszutauschen, sozusagen B3D2.