Re: DogCows or Polymorphism in the Hurd

[Jonathan S. Shapiro]

I find this entire discussion terribly confusing. We have established
that the DogCow style of interface (the one where interfaces may use
multiple inheritance) is an overwhelmingly bad idea from two
perspectives:

  It is semantic nonsense
  It is pragmatic nonsense (causes tremendous unresolvable confusion for
    programs)

Either issue, taken alone, is intrinsically fatal to the entire concept,
and is a sufficient reason to put this design option to death. Nobody
has been able to propose any sensible resolution to *either* of these
issues. Why should we discuss it further?

At the same time, we have another proposal that preserves the good
attributes of idea with essentially *none* of these confusions: the
polytype super-interface. The only issues that have been raised with
polytype concern consistency across views. These are valid issues, and
important concerns in the design of certain multi-view translators, but
they are not intrinsically fatal to the entire concept.

Perhaps there is a third (or a fourth) option that we should consider as
well, but it is time to shoot the poor DogCow and stop its misery. I
would not treat a dog OR a cow this badly. :-)


shap