Re: CASL - V 0.95

[Don Sannella <dts@dcs.ed.ac.uk>]

Michel Bidoit writes:
> BTW, why have we decided that the order of basic-items is irrelevant ?
> What is the rationale for this ??? Are we so sure this is the right
> decision ??? Just makes tools more complex, semantics less clear, etc.

Peter Mosses responds:
> [Good question!  We could require linear visibility, just as in
> libraries - except perhaps in TYPE-DEFN-GROUP, where one has the usual
> problem with mutually-recursive types... --PDM]

I think linear visibility would be an real improvement.  I agree
wholeheartedly with Michel that what we have now just makes everything
more complex.  Having had plenty of experience with linear visibility
in SML and EML, I've never seen a convincing reason to want anything
else, so I don't see what advantage is being bought at the cost of the
increased complexity.

If there is no consensus on this, maybe we could impose linear
visibility and ask people who disagree with this to exhibit examples
between now and April that illustrate why this is the wrong decision.

[Perhaps this change provides a further (slight) motivation for taking
a TYPE-DEFN-GROUP as a SPEC, rather than as a BASIC-ITEM, since it
would otherwise be the only construct breaking the linear visibility
in BASIC-SPECs?  Does anyone see any disadvantages of these changes?
If so, please react straight away!  --PDM]

Cheers, Don

[DON, BERND: Are you happy with the form of your proposals for LOCAL
DEFINITIONS AND DECLARATIONS in App. C?  Should I incorporate them as
they are, assuming nobody raises any problems with them?  It doesn't
seem that they are attracting any comments... --PDM]