🚀 go-pugleaf

   Date:	Tue, 4 Aug 1992 16:11:44 -0700
   From:	jonl%kuwait@lucid.com (Jon L White)

   Should ENSURE-GENERIC-FUNCTION coerce its :METHOD-COMBINATION
   argument?  That is, should it be acceptable to pass 'APPEND to
   ENSURE-GENERIC-FUNCTION rather than the method-combination 
   metaobject?  What do you think?

Well, this is a little hard to answer, since so little of how method
combination metaobjects is defined in the spec.  But, assume for the
time being a straightforward approach where method combination types
name classes of method combination metaobjects (what namespace those
classes are named in is an issue of course).

I believe ENSURE-GENERIC-FUNCTION should be able to accept a method
combination type name, as well as the metaobject.  A further question is
whether it should be willing to accept a list, where the car is the name
and the cdr is initialization arguments (or something like that).

This whole thing is further complicated by the dance we tried to do
around compile-time vs. load time.  It would probably all work much
better if there were separate generic functions called at each time---it
didn't seem like that would be acceptable before, because of the general
concern about not specifying anything to do with the compiler.  I
imagine some people will want to play with some sort of layered protocol
to improve this though.

Gregor