>The problem is that holes at the type level are a difficult feature to
>offer: they require linear types in the compiler. As an 
>optimization, it is a rather high-level one, and maybe not so 
>easy to know when it will apply.

Perhaps, but it's easy for a compiler to offer support
for "tail-allocation" (i.e., a tail-call except for a
constructor application) which is what you need for
a tail-recursive append or map.  Perry Cheng implemented it in
the TIL compiler in about a week if memory serves and
it was a tremendous improvement in performance without
any magic.  

Yasuhiko Minamide wrote a paper on how to model this
well (I think it appeared in ICFP).  The approach used
in our Typed Assembly Language paper is yet another
approach based on a simple subtyping trick with initialization
flags.  It didn't require linear types at all and 
instead of implicit subtyping, you could accomplish
the same thing with an explicit (type-safe) up-cast.

So, all in all, it's quite possible to have the compiler
implement this optimization for the common case of
tail-allocation, and if you think it's more generally
applicable, then you could move to something like TAL's
initialization flags (though I would prefer the former
option.)  

-Greg

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners