[ODE] Infinite sliding... ?

[Jason Perkins]

On 8/18/06, David Walters <hidden.asbestos at googlemail.com> wrote:
> Step alterations should still allow this code to perform all of these
> actions once with minimal fuss. Collision needs to be detected as soon
> as possible, not after X steps.

IMHO this is a good argument against a built-in fixed-step loop. I can
think of other things (like data logging, joint feedback and breaking)
that could end up here as well. Though I'm hoping for an opposing
viewpoint - I would like to see a fixed stepper in the core.

Instead, or at least in the meantime, we can improve the documentation
and samples. Jean has already started on the wiki (thanks!)


> ps. Has anyone had a chance to try my patch from last night?

I haven't applied it, but I did read through it. It looks good and
could open up future optimizations, but it will break everyone's code.
It's not bad - just the call to dWorld...Step() but still it makes me
uncomfortable. How do other people feel about moving the stepsize
argument out of dWorldStep() and into a separate function? We are not
yet at 1.0 so breaking the API is still an option.

Jason