On 18 May 2017, at 10:42, Didier Verna <didier@lrde.epita.fr> wrote:

zbyszek <zbyszek@mimuw.edu.pl> wrote:

Dnia 2017-05-17, śro o godzinie 15:24 -0400, Sam Steingold pisze:

If you are defining the method combination, you have way more freedom
and flexibility than mere before and after. Basically, you can do it
yourself.

But Didier is asking about BUILT-IN method combinations. Possibly it
was hard to define reasonable agreed semantics for before an after
methods in the case of something like AND or APPEND (technical
troubles aside).

Right. I was merely curious. It's pretty obvious to me why you
wouldn't allow CALL-NEXT-METHOD in non-standard built-in combinations,
but I can't figure out why or how before and after methods could be
problematic, so I was wondering...

I’m just guessing, but one reason I can think of is that almost all of the built-in method combinations (except for standard and progn) are applicative. before/after methods don’t have a direct impact on the return value of a generic function call, so their primary purpose is to allow for specifying side effects, which presumably doesn’t make a lot of sense for applicative combinators.