[ODE] Values required to produce at least semi-realistic friction using the default constant-force-limit model?

[Megan Fox]

I've been using the default friction (constant-force-limit, looking at the
docs), and I can't seem to find any value that results in even
semi-realistic friction.  I either set some fraction of dInfinity, and
get -very- sticky objects that barely even respond to objects 300x their
mass (they go flying if I drop the bullet on them, but only barely move if I
roll the bullet into them), or I set a constant to mu (I believe I tried
values up into the thousands), and end up with ice-skating boxes.

Is it just not possible to get reasonable results out of the default
friction?  I'm sure the friction-pyramid implementation is more accurate,
but the docs seem to suggest that it is significantly more expensive (and so
not the the best choice for general case "random box bouncing across a
surface" collisions).


I don't need ultra-accurate results here, this is far from a sim... I just
need to, say, create a room.  The floor is wood, as are the walls, and there
is a counter-top to one side.  The objects sitting on the counter top should
clunk around on the wooden not-horribly-slick-but-not-sandpaper-either floor
a bit if I knock the table over; They shouldn't ice skate, but they also
shouldn't act as if the ground is made of an adhesive/sandpaper blend should
the player kick one after it has fallen to rest.


(and I realize having a bullet 300x as massive is going to break the
system - I would never have something that massive in the final product - I
was just trying to get the dratted objects to move when the bullet rolled
into them, with largely no success.)

Thanks,

-Megan Fox
Lead Developer, Elium Project (http://www.elium.tk)