Pierre Weis wrote:
>              Relational types in Caml
>   
Very nice indeed.  As a matter of fact, I was just reading (and 
enjoying) the ESOP paper in the plane yesterday!

My question is: why is this done as a pre-processor instead of as an 
extension of metaocaml?  Certainly some of the features of Moca can be 
implemented in metaocaml (I have done this myself for rings and 
monoids).  The added advantage of using metaocaml is, of course, the 
types.  As a (syntactic) pre-processor, getting good error messages out 
of Moca will be rather challenging, no?

Jacques

PS: Genuine question, not criticism.