On Sun, Dec 10, 2006 at 11:20:51AM -0500, D'Arcy J.M. Cain wrote:
> On Sun, 10 Dec 2006 16:13:07 +0100 (CET)
> Hubert Feyrer <hubert@feyrer.de> wrote:
> > On Sun, 10 Dec 2006, Jukka Salmi wrote:
> > > While changing the third-party software to include sys/types.h fixes
> > > the problem, I think it's NetBSD's fault: net/pfil.h should know about
> > > `u_long' by including sys/types.h, shouldn't it?
> > 
> > IIRC we've had this discussion a few times before.
> > IIRC the concensus (decree?) was always that applications should know what 
> > headers are needed for a certain (other) header, to not pull in things 
> > that my not be needed.
> > 
> > (And no, I never understood that either, and always thought headers should 
> > pull in what they need, but that's just my personal, uneducated opinion)
> 
> I agree with you, Hubert.  Third party code should pull in any headers
> that they need but they shouldn't need to have inside knowledge of what
> those headers use.  After all, it's not like we actually save
> anything.  If the header needs something, someone needs to pull it in.
> Why not take responsibility for its own requirements?

Possibly to avoid namespace pollution.

	-is
-- 
seal your e-mail: http://www.gnupg.org/