Subject: Re: a new KNF (and some comments)
To: Martin Husemann <martin@rumolt.teuto.de>
From: Jason Thorpe <thorpej@nas.nasa.gov>
List: tech-kern
Date: 01/22/2000 09:17:47
On Sat, 22 Jan 2000 17:10:18 +0100 
 "Martin Husemann" <martin@rumolt.teuto.de> wrote:

 > > In headers files that include user-accessible prototypes, you CANNOT
 > > include variable names, as they intrude on the user's namespace.
 > 
 > I buy this, but it's a shame. Seems like a constructed argument to me, I've
 > never been hit by such a thing in everyday work. That was the reasons K&R
 > suggested macros to be all uppercase?

It's a perfectly valid argument, and I *have* been nailed by this before,
though not on a NetBSD system.

 > > Documentation documents the interface.  Header files are for the compiler
 > > to read.
 > 
 > Here, here! And NetBSD is a great example to what level this works ;-}
 > Sorry, couldn't resist.

Just because our kernel interface documentation is lacking doesn't mean
that we should pollute the headers files with incomplete documentation.
(Sure, you know the arguments, but what does the function do with them?)

        -- Jason R. Thorpe <thorpej@nas.nasa.gov>