Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!watmath!clyde!burl!mgnetp!ihnp4!zehntel!hplabs!sri-unix!Kahn.PA@XEROX
From: Kahn.PA%XEROX@sri-unix.UUCP
Newsgroups: net.lang.prolog
Subject: Cohen's arrays
Message-ID: <12372@sri-arpa.UUCP>
Date: Fri, 3-Aug-84 20:40:00 EDT
Article-I.D.: sri-arpa.12372
Posted: Fri Aug  3 20:40:00 1984
Date-Received: Wed, 15-Aug-84 06:43:01 EDT
Lines: 13

While I found Cohen's code hard to read it did seem to have the
advantages he claims and seems to work.  It is hairy and I would
like to see  some English describing  how it works.  I should
point out that "Kahn's algorithm" was jointly developed by Lars
-Henrik Eriksson and Manny Rayner and myself.  Its true that in
general the chain of changes needs to be copied but in what we
think is the most common use of arrays (linear chains of updates)
then this chain to be copied is empty.  It seems that  an update
in Cohen's scheme costs about 8 words vs. 3 or 4 for us but I don't
think this is so important.  I'm looking forward to seeing  a clear
description of Cohen's algorithm since it seems so promising.

-- Ken Kahn