RESTIVO@SU-SCORE.ARPA (03/28/85)
From: Chuck Restivo (The Moderator) <PROLOG-REQUEST@SU-SCORE.ARPA>
PROLOG Digest Thursday, 28 Mar 1985 Volume 3 : Issue 15
Today's Topics:
Editorial - Request
Implementations - Cuts & RF-Maple & CP Semantics
----------------------------------------------------------------------
Date: Wed 27 Mar 85 11:35:29-PST
From: Fernando Pereira <Pereira@SU-CSLI.ARPA>
Subject: Short and to the point
The current trend of longer and longer contributions
in the Prolog Digest makes it almost impossible to
follow the discussions. I find myself forced to skip
of just skim any contribution longer than a couple of
screens.
Here are some suggestions on how to improve the
readability of contributions:
1. Keep each contribution within the size of a single
typewritten page (~60 lines).
2. Don't discuss more than one topic in a contribution;
if you want to address several topics, make separate
contributions with separate subject lines.
3. Use an editor to compose your contribution, and edit
it before sending; this will help cut down repetition.
4. Abstain from ``ad hominen'' arguments, innuendo,
tasteless irony and other such flamage; readers looking
for such material are much more likely to find enjoyment
in the satirical literature in their local library.
5. If your argument REALLY needs more than 1000 words,
WRITE a paper instead and submit it to a conference or
journal. If your idea is worth more than 1000 words you
owe it to yourself to make it available to a wider audience
that the readership of this Digest.
6. Reread and reread your contribution: there's no
point in making enemies unnecessarily.
-- Fernando Pereira
------------------------------
Date: 27 Mar 85 17:41:48 +1000 (Wed)
From: decvax!mulga!mungunni.oz!lee@Berkeley
Subject: read only annotations
I agree with Vijay Saraswat's opinion that !
annotations are nicer than read only variable
annotations (?). I like some sort of mode
declaration even more.
The initial proposal for ? was more 'powerful'
than ! or modes, in particular, the ability to
pass around chains of read only references.
In order to make ? more easy to understand and
implement, its power has been reduced. However,
mode declarations are still conceptually much
simpler. I suspect they are easier to implement
also, especially if you are trying to compile
things as much as possible. After all, its the
heads of clauses which are compiled most, whereas
? occurs mostly in calls.
Does anyone know if the latest version of ? has
any more power than ! or modes? If so, is this
extra power useful in practice?
-- Lee Naish
------------------------------
Date: Wed 27 Mar 85 17:09:57-MST
From: Uday Reddy <U-REDDY@UTAH-20.ARPA>
Subject: RF-maple
Paul Voda from V3 #14:
"the claim Uday makes that without unification
there cannot be output variables (values?) is
rather strong".
I claimed no such thing. FORTRAN produces output
values and does'nt use unification.
My question was "Can RF-Maple solve?". It has'nt been
answered. From all that Paul says, it appears that it
can't. May be it is interesting in spite of that. He
can also choose to call it a "logic language". I have
no objection to that. But, then, it should be kept in
mind that his notion of a "logic language" is different
from the kind of logic languages we are discussing here
and that there is no basis for comparing it with these.
-- Uday Reddy
------------------------------
Date: Wed 27 Mar 85 17:53:17-MST
From: Uday Reddy <U-REDDY@UTAH-20.ARPA>
Subject: Semantics of CP
In response to Paul Voda, V3 #14:
1. What is the meaning of the goal p(Y) in the context
of the clause
p(X) :- var("X"), X=1.
with the proposed "" construction?
2. Just as I don't agree with purity-as-God-given
principle, I don't also agree with classical-logicians-
didn't-do-this principle. Even if I have to "escape to
metatheory" or the vital connection to the predicate
logic is lost, maybe I can still live with it - as a
programmer. But, the level of complexity I can com-
fortably live with - is bounded.
3. "Is it not significant that whenever we have control
different from the head normal reduction [normal
order reduction?] the denotational semantics is in
trouble?"
Not true. The reduction order depends on whether the
functions are strict or non-strict. But, if different
reduction orders are used for the same function,
denotational semantics would naturally have trouble
with it.
4. "Operational semantics [of CP] can be explained by
a formal theory ..."
Sorry. Operational semantics does'nt "compose", no
matter how formal a theory. I want to know the
meanings of higher-level concepts in terms of its
constituent lower-level parts.
"Computations are proofs".
Exactly. That is why a proof theory is not enough.
It is no better than "run-it-and-see". Proof theories
CAN be compositional, like for instance Hoare-Dijkstra
semantics for imperative languages. But, not any proof
theory is compositional.
-- Uday Reddy
------------------------------
End of PROLOG Digest
********************