Christophe TROESTLER wrote:
> 
> Wouldn't it be better to have a compiler switch (say "-w T/t", off by
> default) to overide that behavior (new definitions override old ones
> as with [open])?  Of course that makes it impossible to write M1
> signature without aliasing [t] and [Std] but IMHO that's fine.

Note that OCaml's type system fundamentally relies on the fact that type 
names cannot be shadowed in structures.

> This
> way, no new syntax needs to be introduced (and one is not forced to
> alias modules one does not care about -- e.g. utilities).
> 
> module M = struct
>   include M1
>   type m1_t = M1.t
>   module M1Std = M1.Std
> 
>   include M2
>   type m2_t = M2.t
>   module M2Std = M2.Std
> 
>   module Std = struct
>     include M1.Std
>     include M2.Std
>   end
> end

While your proposal probably could be made to work in this particular 
case - the compiler had to figure out that it can break the dependency 
on the first t by reordering m1_t relative to the fields from M1 in the 
derived signature and substituting m1_t for t in these fields, likewise 
for the other elements - it is *far* from obvious if and how this can be 
inferred in general, particularly when your modules go higher-order.

I believe it would require something akin to join and meet operations on 
signatures, but it is known that these don't exist for some very similar 
systems (signatures for higher-order modules don't form a lattice).

Cheers,

   - Andreas

-- 
Andreas Rossberg, rossberg@ps.uni-sb.de

Let's get rid of those possible thingies!  -- TB