[ODE] Pitfalls of using GeomTransformGroup?

[Megan Fox]

I'm considering using GeomTransformGroup as a simple means of
implementing  both ragdoll and the generic attachment of shapes to a
pre-existing model (for instance, attaching a sword to a ghost-body
ragdoll, without the logic animating the ghost needing to know
specifically about the sword - the same code could attach armor plate
collision bodies, shields, and whatever else).

Now it at least appears that this would work out excellently, I'd have
all the ragdoll positioning data in bone-relative terms ready to poke
back into my animation code and so on - but surely there's some (good)
reason why this hasn't made core yet?

(the other alternative is, of course, to do the conversions between
bone space and object space myself when communicating data back and
forth, brew up a little logic that special cases "attached" geoms vs
pre-existing geoms that makes them easy to add and remove run-time,
etc - not a big deal really, but if GeomTransformGroup works well, no
point in doing all that myself)


Thanks,
-Megan Fox