Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!ames!attctc!pollux!ti-csl!m2!gateley
From: gateley@m2.csc.ti.com (John Gateley)
Newsgroups: comp.lang.scheme
Subject: Re: Request for Comments: A new n-ary function construction
Message-ID: <99631@ti-csl.csc.ti.com>
Date: 27 Nov 89 03:36:38 GMT
References: <891122-121343-4098@Xerox>
Sender: news@ti-csl.csc.ti.com
Reply-To: gateley@m2.UUCP (John Gateley)
Organization: TI Computer Science Center, Dallas
Lines: 24

In article <891122-121343-4098@Xerox> Pavel.pa@XEROX.COM writes:
>[An interesting idea for handling variable numbers of arguments]

I like the idea, though it is a large change (if it is implemented
efficiently). I do have one point to make against it: the intertia
of the list data type. Not only is there apply, but also map(car) and
for-each, (and if you like CL, mapcan, mapcon, map-the-world :^).
I would like a clean integration with some sort of looping structure,
but this might mean (re)designing a MIT style LOOP macro. Any ideas?

>[Note: I'll point out right here that nobody I've shown this to has liked
>having strings in the lambda-list.  It just seemed better to me than
>&arbitrary or #!arbitrary.  Suggestions are solicited.]

After some thought, strings don't sound so bad to me. Consider:
identifiers are really objects which we know at compile time to be
symbols. (Is this true in Scheme? Do there exist identifiers which
have no symbol or vice versa? I've been using CL too long). Along
the same lines, we could have C-Strings (compile-time strings) which
appear in lambda lists. The only significance to the name "C-string"
is that identifiers and symbols have different names.

John
gateley@m2.csc.ti.com