Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!husc6!uwvax!rutgers!rochester!ken
From: ken@cs.rochester.edu (Ken Yap)
Newsgroups: comp.unix.questions
Subject: Re: spell bug?????
Message-ID: <1262@sol.ARPA>
Date: Tue, 11-Aug-87 13:12:43 EDT
Article-I.D.: sol.1262
Posted: Tue Aug 11 13:12:43 1987
Date-Received: Thu, 13-Aug-87 03:23:57 EDT
References: <541@augusta.UUCP> <727@houxa.UUCP> <1673@ncr-sd.SanDiego.NCR.COM>
Reply-To: ken@cs.rochester.edu (Ken Yap)
Organization: U of Rochester, CS Dept, Rochester, NY
Lines: 17
Keywords: is this in every system?

|A workable, but kludged, solution -- for sites with only the BSD
|"spell" -- is to add these strings, and any other known nonsense
|words, to the local stoplist used by the spell-checker.  The
|real solution, of course, is to fix the source (for those sites
|that have source).  (Is anyone in Berkeley listening? ;^))

Nobody has mentioned the possibility that the words in question fell
through spell's probabilistic detection algorithm. Last time I looked
at spell, it hashes each suspect word with 12 (?) independent hash
functions to get all 12 bit addresses and checks to see if all bit
locations were turned on in the bitmap. If they all are, the
word is let through as legitimate.

Maybe the bitmap is too full or the hash doesn't work properly.  Only
in the second case would the source be fixable.

	Ken