Re: CLOS-style MOP for Guile ?

[Mikael Djurfeldt <mdj@nada.kth.se>]

mike@olan.com (Michael N. Livshin) writes:

> Here is what I consider proper generic <-> module system interaction:
> 
> (define-module (my-module)
>  #:use-modules (used1) (used2))
> 
> (define-generic-method some-known-generic (<myclass>...)
>   (do-stuff))
> 
> 1. In (my-module), some-known-generic has a method for <myclass> and
>    also has all the methods that were added to it in (used1) and (used2).
> 
> 2. In (used1) or (used2), some-known-generic is *not* extended for <myclass>.

While I agree that your suggestion is the most common sensical one, it
seems to mix badly with some fundamental concepts in Scheme.

In Scheme variables are bound to locations, nothing else.

The module system is only responsible for supplying mappings between
variables and locations.

What we store in those locations is an entirely separate issue.

If we have stored a generic function object in a variable in the
module `used1' and the variable binding is exported, that means that
whenever we add methods to the object, those methods will also be
visible in the module `used1'.

I guess the most reasonable way of dealing with this is that the
module system is modified so that it not only copies bindings when
exporting variables, but also makes a copy of the object stored in the
location of the exported binding if that object is a generic function
object.

/mdj