From kde-core-devel  Fri May 20 08:17:00 2005
From: christian.loose () hamburg ! de
Date: Fri, 20 May 2005 08:17:00 +0000
To: kde-core-devel
Subject: Re: Re: QUrl in KDE 4
Message-Id: <12432494.1116577020693.JavaMail.meinbereich () kalifornien>
X-MARC-Message: https://marc.info/?l=kde-core-devel&m=111657704119815

-----Ursprüngliche Nachricht-----
Von: jfirebaugh@kde.org
Gesendet: Freitag, 20. Mai 2005 05:08
An: kde-core-devel@kde.org
Betreff: Re: QUrl in KDE 4

> On May 19, 2005, at 7:02 PM, Frans Englich wrote:
> > As Thiago, I disagree, because with that approach you will find the  
> > problem in
> > any object oriented design; where something is a subset,  
> > constrainment, of
> > something wider -- what class hierarchies usually are about,  
> > AFAICT. A square
> > is a subset of the recangle's value space, so to speak.
> [snip]
> > Hence, I don't see representation problem Thiago sees. Perhaps an  
> > elaboration
> > can be provided?
> 
> On the contrary, class hierarchies are ill suited to this type of  
> subset relationship:
> 
[snip]
> 
> Inheriting the other way around doesn't work either.
> 
> See also http://c2.com/cgi/wiki?CircleAndEllipseProblem
> 
> The URL/URI relationship is very similar.

One can find another good explanation for this problem in the C++ FAQ Lite.
See [21.6] till [21.12]: http://www.parashift.com/c++-faq-lite/proper-inheritance.html#faq-21.6