Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/18/84; site arizona.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!mhuxr!mhuxb!houxm!ihnp4!arizona!gary
From: gary@arizona.UUCP (Gary Marc Levin)
Newsgroups: net.nlang
Subject: Re: Semantic Reversals
Message-ID: <22060@arizona.UUCP>
Date: Mon, 25-Feb-85 14:57:07 EST
Article-I.D.: arizona.22060
Posted: Mon Feb 25 14:57:07 1985
Date-Received: Wed, 27-Feb-85 08:01:39 EST
References: <> <258@unm-cvax.UUCP> <486@ptsfa.UUCP>
Organization: Dept of CS, U of Arizona, Tucson
Lines: 20

> One curious semantic reversal occurs in the expression:
> 	The exception proves the rule.
> Originally 'prove' used to mean 'test' (cf. 'probe', a related word).
> But as the meaning of this word changed, the expression changed also.
> Until I heard of this explanation I [mis]understood this expression
> to mean:
> 	Don't worry about the rule at hand having an exception
> 	since all rules have exceptions, here's the one exception that
> 	qualifies this rule as a true rule.
> Rob Bernardo, Pacific Bell, San Francisco, California

Actually there are two reversals in the phrase.  "exception" only
meant "unusual", not "not covered".  The sentence really is a reminder
to use boundary cases when testing programs.

(I seem to remember going to a lot of effort once to find the history
of all this.  It may just be folk-etymology on my part.  Still, makes
more sense than the usual "modern" reading of the sentence.)
-- 
Gary Levin / Dept of CS / U of AZ / Tucson, AZ 85721 / (602) 621-4231