From: utzoo!decvax!ucbvax!C70:info-cpm
Newsgroups: fa.info-cpm
Title: Fast spelling checker
Article-I.D.: ucb.1733
Posted: Tue Aug 10 00:31:51 1982
Received: Thu Aug 12 04:59:58 1982

>From FONER@Mit-Mc Tue Aug 10 00:30:27 1982
Yes, that sounds fair.  The worst case for such an algorithm occurs
when every word in the input file is unique, which would necessitate
essentially a copy of the input file to disk if the file itself is
larger than your physical memory (losing all the time because of
the necessity to maintain the output sorted alphabetically...  even a
binary tree would necessitate a lot of head motion, I fear).

However this is a worst case, and you could always simply store only
what won't fit in memory on disk.  Probably you will usually find that
you can hold the whole alphabetized list in memory, and thus win.
Short files (smaller than physical memory) will always win.  Large
ones might be profitably split up if the chances of unique words
filling memory is high, but this is probably rare...  almost any given
document will not be using very many \unique/ words unless you are
checking a dictionary or Brown's Corpus or something equally
inappropriate.

Good luck!