[ODE] dCylinder problems and future

[Adam D. Moss]

John Miles wrote:
> I don't see any reason not to go ahead and add dcylinder.cpp to the main
> library, as long as its limitations are mentioned prominently in the docs.

I'm not saying that this will or won't happen, but I wanted to
point out that adding a geom to core ODE isn't something that
should be done lightly.  Every geom that gets added means
increasingly more work to add any other future geoms, since
all core ODE geoms should know how to collide with all other
core ODE geoms and all of those collision cases have to be
written.

A dCylinder geom (though not necessarily this implementation
in its current state) is IMO quite a good candidate for
the core because it's useful in itself and can only very
roughly be approximated by the other primitives -- if it went in,
the CCylinder as a low-level collision geom could in theory be
retired and perhaps replaced by a fake (though the CCylinder is
somewhat more optimal than a cylinder-and-two-spheres, if a
CCylinder is really what you need), then future geoms would only
need to know how to collide with a sphere and a cylinder...

Maybe in the future some space-cadet will submit code to allow
geoms to be constructed and collided via a CSG hierarchy, so for
example a cylinder trivially becomes a CCylinder bounded by two
planes. :)

--Adam
-- 
Adam D. Moss   . ,,^^   adam at gimp.org   http://www.foxbox.org/   co:3
"At this point the rocket becomes engorged with astronauts."