Subject: Re: MSGBUFSIZE exposed to user-land
To: Gordon W. Ross <gwr@mc.com>
From: Darren Reed <darrenr@cyber.com.au>
List: tech-kern
Date: 02/19/1998 08:19:27
In some mail I received from Gordon W. Ross, sie wrote
> 
> > Date: Tue, 17 Feb 1998 06:54:14 -0800 (PST)
> > From: David Brownlee <abs@anim.dreamworks.com>
> > 
> >     Question to tech-kern:
> > 
> > 	Nothing in userland uses the MSGBUFSIZE definition in 
> > 	machine/param.h (since it can very from kernel to kernel).
> [...]
> > 	Could/should the machine/param.h MSGBUFSIZE definition be 
> > 	protected in #ifdef _KERNEL?
> 
> >     Reply to jspath:
> > 
> > 	I doubt fetchmail is using the MSGBUFSIZE define for the same
> > 	purpose, so that would not be the desired behaviour.
> > 
> > 	Some options are
> [...]
> > 	   b) Modify <machine/param.h> to hide MSGBUFSIZE from userland.
> 
> There are many, many definitions exposed to user-land that can be
> considered namespace pollution.  MSGBUFSIZE is just one of them.
> 
> Fixing that would a fairly big project.
> Is this something that should be fixed?
> Is anyone willing to work on it?

Consider that not only do many NetBSD programs potentially need fixing, but
much 3rd party source that has relied on things being "that way" for years
may also end up needing "fixed".

I think the adage "if it's not broken, don't fix it" should apply here.

Darren