This is the mail archive of the
kawa@sources.redhat.com
mailing list for the Kawa project.
[bdc@zurich.ai.mit.edu: Re: string srfi]
- To: <per at bothner dot com>
- Subject: [bdc@zurich.ai.mit.edu: Re: string srfi]
- From: shivers at cc dot gatech dot edu
- Date: Thu, 8 Mar 2001 17:03:52 -0500
- CC: <nferrier at tapsellferrier dot co dot uk>, kawa at sourceware dot cygnus dot com
- Reply-to: shivers at cc dot gatech dot edu
C'mon, guys, be serious. I'd put that on my "clean this up when we've
absolutely run out of useful things to do or someone in a position of
authority complains seriously about it " list...
If you really, really want to waste time on this, I just went & checked. The
four remaining functions that came from MIT Scheme are
%STRING-{PREFIX,SUFFIX}-LENGTH{,-CI}.
If you decide to get a friend to describe the API and rewrite them clean-room,
and you mail me source-code and make it either public domain or BSD, I will
gratefully receive them, make my own pass over them, and stick the results
into the reference lib, cite you as a contributing hero, and rip out the MIT
copyright.
But you must write these things *tight*. They are the inner loop of
all string comparisons; they were microcoded on the early MIT Schemes.
Tight iterations, fast-path cases for EQ? strings, do it right.
-Olin
From: Per Bothner <per@bothner.com>
To: "Nic Ferrier" <nferrier@tapsellferrier.co.uk>
Cc: kawa@sourceware.cygnus.com
Subject: Re: string srfi
Date: 26 Feb 2001 19:19:26 -0800
"Nic Ferrier" <nferrier@tapsellferrier.co.uk> writes:
> Is there any interest from anybody in seeing a Kawa implementation of
> the string srfi? (by olin shivers).
It would probably be desirable. I am concerned about the license of
the reference implementation. I don't think clause 2 or 3 of the
MIT Scheme license (which covers part of the code) are compatible
with the GPL, so cannot be distributed with Kawa. However, a good
part of the implementation, including the KMP string-search code
(I guess the hardest part of the code) is covered by the scsh
license, which I don't see any problems with. Hopefully, Olin
will be willing clarify which parts of the code are contaminated
by the MIT Scheme license, so those can be re-implemented.