[Prev][Next][Index][Thread]

Re: semantics



Neil has some very good points about sharing semantics between
CA systems. OpenMath is attempting to solve this problem, so, in some
sense, we don't have to do so here. But I think we do need to allow our
markup to be semantically annotated. To me, this means allowing two
possibilities:

1. type information on expressions
2. context information on expressions

Here a type annotation on sin(x) might say "sin(x) is an object of type
Expression with Integer coefficients". The context annotation would
say "the sin function mentioned here is the one from the Trigonometry
context." The actual type or context in this example is irrelevant.

I see this committee as mainly a language design group. I think we need
to support both of the above forms of annotation WITHOUT REGARD TO
THE TYPES OR CONTEXTS ACTUALLY USED. That is, we let users
specify types and contexts but we don't define them. We let OpenMath
define the contexts, say, and we let AXIOM and Cayley fight out the
type names they wish to share. Of course, we might want an attribute
somewhere near the beginning of an expression stating that, say, the
OpenMath context system will be used in what follows.

Bob


Follow-Ups: