Francisco J. Lopez Fraguas schrieb:
> In any case, the use of generators/rewriting to replace free variables/narrowing
> does not need any new theory nor formal setting (this doesn't mean that
> nothing interesting can be investigated).
> Existing formalisms (CRWL and its
> rewriting-with-let-bindings counterpart --see our ppdp07--, or the operational
> semantics of JSC mentioned by Michael) cope perfectly with generators,
> which conceptually are not different from any other function, as happens
> also with the function ?.
Yes, I fully agree. And that is the merit of such an approach that you
do not need any new concepts but rather a smaller subset of the original
concepts. And what is interesting to investigate (as you also stated) is
if this smaller subset can be extended in a simple way such that one
regains the advantages of free variables (like generality of solutions)
> I don't think that giving to generators or to ? a special
> status from the point of view of the semantics (intended or formal) is a real
> need,
> not even an advantage.
Yes and know, depending on what semantics you think. From a denotational
point of you I fully agree with your statement. Again this is the
advantage to have well studied concepts. But for the strategic,
operational point of view it makes much sense to treat (?) specially.
Remember that Sergio said that he is looking into possibilities for
parallelizing strategies. When thinking about that it is a great help to
know that (?) combines independent computations. And any sharing across
(?) essentially says that the computations may not be so independent
after all.
> A different question is the interest of developing an operational procedure
> where computations produce set of values, instead of individual ones.
>
> This is interesting in many aspects, has been addressed several times,
> and might profit of a sage usage of generators combined with ?, as it seems
> you are working out in your recent papers.
> But I don't think that this needs changing the foundations of FLP.
> Instead, in my opinion such collecting procedures should guarantee
> equivalence with the semantics given to individual computations in previous
> proposals (CRWL, JSC, ppdp07).
> This was done for instance, for CRWL and the set-valued semantics in our
> lpar01/flops02 papers.
But even if such an approach to "collecting procedures" guarantees
equivalence with the semantics for independent computations, it is still
a substantial extension of he foundations of FLP, or don't you think?
And I have the intuition (maybe with Sergio ?) that studying the sharing
structure and having the strategy give special treatment to (/) might be
a key issue for that.
Greetings
Bernd
_______________________________________________
curry mailing list
curry_at_lists.RWTH-Aachen.DE
http://MailMan.RWTH-Aachen.DE/mailman/listinfo/curry
Received on Di Nov 06 2007 - 18:14:06 CET