🚀 go-pugleaf

In article <9235622.25712@mulga.cs.mu.OZ.AU> fjh@munta.cs.mu.OZ.AU (Fergus James HENDERSON) writes:
|A ptrcmp that always returned 0 would be useless, IMHO.

Such an implementation would allow people to write code that does the
ptrcmp.  If the ptrcmp is zero, then the pointers could be compared.
If the pointers didn't compare equal, then the invoking code would know
that this machine doesn't support ptrcmp, and then the invoking code would
fall back on a secondary, possibly slightly slower [but more globally 
implementable] approach -- such as comparing object surrogates.

Thus such a ptrcmp would allow "optimal" performance on those systems
that can reasonably implement a total ordering, and would allow near
optimal performance on those machines that can't reasonably implement
a total ordering.  Rather than insisting that those systems that can't
reasonably implement ptrcmp do so anyway, or alternately not implement it
at all, leading to "portable" code that isn't in practice portable anyway.