Generative record types

[Shriram Krishnamurthi <shriram@cs.rice.edu>]

>			    Unfortunately the proposal that you make doesn't
> immediately seem very different to me from the `make-record-type' proposal
> that we have -almost- agreed to on several previous occasions.

Chances are there are no substantive differences, only (mostly)
syntactic ones.  Perhaps define-struct can be defined as a macro on
top of make-record-type; Richard Kelsey once kindly provided me with a
macro for define-structure atop define-record-type.

>								  How does
> this proposal address the concerns that have caused `make-record-type' to
> repeatedly fail to pass?

Perhaps it would help me and others who have been on the sidelines if
someone (you?) could summarize why those proposals failed in the past
(and how, if at all, those proposals differed from ours).  My
impression, gleaned from sampling past discussions, is that a lot of
discussion centered around issues like opacity.  We tried to
explicitly deflect attention from that.

Also, the condition datatype situation could, possibly, actually
extend the type structure of Scheme.  Hence, perhaps this is a
particularly good time to consider means for giving the user the power
to do this too.  If, in the bargain, we can agree on a syntax for
records ("structures", as some of us call them), so much the better.
I don't know of a compelling reason why the two cannot be conflated.

Shriram &c.

PS: Interestingly, I believe the exception type is the one place in
    SML where a datatype is extensible.