"Disjoint" relation
I agree with Arnold that the "Disjoint" relation deserves to be a
primitive. Interestingly, Motion 21 also requires synthesizing it
out of two of the "overlapping" states.
It seems that any attempt to base the selection of primitives that is
based on some abstract underlying structure (be it BRA, reflection,
or overlap type) will leave out some practical predicates because they
can be derived "easily" from the generators.
Some predicates are difficult to standardize because several variants
are conceivable (and perhaps even in use), but others (like Disjoint)
have a clear neough definition, and practical importance, that they
should be added anyway.
It is true that (as Dominique and others have observed) that in many
cases of compound operators at the source program level an optimizing
compiler can derive the simpler and direct code sequence automatically,
but even there this depends on what machine primitives are available.
And, as I've mentioned before, this also assumes built-in support for
Interval Arithmetic, which may take time to become common enough.
Michel.
---Sent: 2010-10-04 13:53:20 UTC