[bitc-dev] Value types and the Value restriction

[Jonathan S. Shapiro]

Swaroop:

I think that your approach on this, which you described in:

  http://www.coyotos.org/pipermail/bitc-dev/2006-July/000742.html

makes sense.

It would be useful if you could send out a followup note that follows
the following outline, so that everybody shares a coherent picture:

  I. The Problem

     Give an example of something that is screwed up under the
     current type rules. Explain why it's a problem.

  II. The Proposal

     Say what the plan/solution is. What are the new rules?

  III. Consequences

     Give some examples that will produce different answers
     under the old and new rules, so that people can see them.


Finally, can you either define what you meant by "contrapositive
position" or perhaps use "contravariant position" (which is what I think
you meant). Note that contravariance has a weird odd/even inversion for
function parameters, so if you mean contravatiant that's great, but you
should check whether what you mean is "argument position". For example,
are there also issues in covariant position?


shap