So far I have found quite a lot of differences between how different packages treat boundary edges. Comparing, say, Modo and Max, I’ve noticed some significant differences in some situations. I can’t remember off-hand what the original paper said about these border cases – I’ll have to have another read.

Certainly, tools like vertex weight maps and edge creasing are completely inconsistent among the variety of pacakges which support Catmull-Clark surfaces – hence the reason why it’s often difficult to transport sub-d surfaces from one package to another. I’ve heard that Lightwave’s surfaces are particularly good, though I’m not sure how ‘good’ was measured.

I’ve been meaning to do some more research into subdivision surfaces in general now that DirectX 11 cards (with hardware tessellation units) are starting to make their way to the shops. I’ll let you know if I find something relevant.