Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!usc!apple!cep
From: cep@Apple.COM (Christopher Pettus)
Newsgroups: comp.lang.c++
Subject: Re: Public vs Private header files in C++
Message-ID: <53598@apple.Apple.COM>
Date: 3 Jun 91 22:22:50 GMT
References: <5243@servax0.essex.ac.uk> <1991Jun3.182909.21539@agate.berkeley.edu> <5tdHm72m@cs.psu.edu>
Organization: Apple Computer, Inc.
Lines: 23

In article <5tdHm72m@cs.psu.edu> schwartz@groucho.cs.psu.edu (Scott Schwartz) writes:
>
>jbuck@forney.berkeley.edu (Joe Buck) writes:
>   Stroustrup said something about this: the member access system is designed
>   to protect against accidents, not maliciousness.  
>
>An then on page 253 gives an example of language features designed to
>prevent maliciousness (noninheritance of friendship), using an example
>named "class Spy".

Stroustrup's choice of words in this example is exceptionally poor, but
the issue is the same: it would be easy to violate the purpose of
protected and private if friendship was inherited (at least that's his
point; IMHO, it's being overly conservative).

The _real_ reason is, of course, the same: without full compiler access
to the class definition, it's size could not be determined, which
violates the classes-as-primitive-types goal of C++ (which is a worthy
one).
-- 
Christopher Pettus -- Object-Based Systems -- Apple Computer, Inc.
MS 3-PK -- (408) 974-0004 -- cep@apple.com -- Link CHRISTOPHE
"Don't let your .h write no check that your .c can't cash."