Newsgroups: comp.lang.smalltalk
Path: cantaloupe.srv.cs.cmu.edu!europa.chnt.gtegsc.com!howland.reston.ans.net!swrinde!cs.utexas.edu!utnut!info.physics.utoronto.ca!ie.utoronto.ca!news
From: "bigg.ie Macintosh netscape browser" <golovch@ie.utoronto.ca>
Subject: Re: 
X-Nntp-Posting-Host: bigg.ie
Message-ID: <D9K3q6.IGC.0.ie@ie.utoronto.ca>
Sender: news@ie.utoronto.ca (USENET News Administrator)
Organization: Dept. Of Industrial Engineering, University Of Toronto
References: <3qga8f$d2t@pipe6.nyc.pipeline.com> <3qkbci$17u@isnews.is.s.u-tokyo.ac.jp>
Date: Fri, 2 Jun 1995 17:39:39 GMT
Lines: 30

jeff@is.s.u-tokyo.ac.jp (Jeff McAffer) wrote:
>
> "Private", by my definition, means the receiver can only be self.  So
> the "correct" test would be (self sender == self).  It's not clear to
> me why one would want to do this sort of check.  Presumeably you have
> designed your system/interfaces correctly and are using them
> correctly.  If not, this might be useful in a testing phase but I
> would be sceptical of shipped code with containing this sort of stuff.
This interpretation of "private" bothers me a bit because it
precludes the use of that method by instances of subclasses 
of the receiver's class. This seems to me to be a rather harsh
restriction, particularly if subclasses are not designed at
the same time as the class in question. 

This brings up the issue of the relationship among subclasses:
Are subclasses logically distinct entities that are entitled
to their privacy, or are they merely a convenient factorization
tool and therefore have not claim to privacy of behavior with 
respect to their inheritors? (They certainly don't have any
formal privacy with respect to instance variables.)

Comments?

 2
G

--
<a href="http://anarch.ie.utoronto.ca/people/golovch/">Gene Golovchinsky</a>
Human Factors
University of Toronto
