Toward a Calculus of Signatures
The designers of languages such as Standard ML and Objective Caml have
paid much attention
to mechanisms for building modules from other modules.
Both the underlying semantics and the expressive power have been well
It is surprising then, that the the mechanisms used to build
interfaces (signatures) from other interfaces are so impoverished.
Signatures languages could be improved by borrowing
ideas from the design of abstract data types, especially of sets.
This note presents
a few examples from real programs, which may convince readers that the
shortcomings are real and that the proposed improvements really would
The note closes with a brief syntactic proposal, which has not yet
been fleshed out with semantics.
The paper is available as
US Letter PostScript (243K) and
US Letter TeX DVI (62K).