> .mli files are highly redundant. Almost without exception all, or at
> the vast majority of .mli information can be generated from the
> underlying .ml implementation. We have programming languages to reduce
> redundancy, not increase it. [...]

We probably don't use interfaces (.mli files) in the same way.

Personally, I write the interface  *before* the code. It clarifies the
design  of  an  implementation,  helping  you focusing  first  on  the
features of  the module, independently of  the way you  will write it.
Therefore, there is no way of  generating the .mli from the .ml, since
it comes first.  Moreover, I don't put the  same kind of documentation
in  the   interface  and  in  the   code:  in  the   interface  I  put
specifications, mainly, and in the code I put implementation details.

In the  extreme situation where there  is no real need  for writing an
interface, you can either simply not write one (this is not mandatory)
or generate it from the code with "ocamlc -c -i".

But writing  an interface is always  a good idea,  since it encourages
you to  abstract data types and  to provide only  small and orthogonal
sets of functions, which cannot be done automatically in most cases.

-- 
Jean-Christophe Filliatre    
  Computer Science Laboratory   Phone (650) 859-5173
  SRI International             FAX   (650) 859-2844
  333 Ravenswood Ave.           email filliatr@csl.sri.com
  Menlo Park, CA 94025, USA     web   http://www.csl.sri.com/~filliatr