From owner-freebsd-hackers  Sun Apr 20 23:35:47 1997
Return-Path: <owner-hackers>
Received: (from root@localhost)
          by freefall.freebsd.org (8.8.5/8.8.5) id XAA19495
          for hackers-outgoing; Sun, 20 Apr 1997 23:35:47 -0700 (PDT)
Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.19])
          by freefall.freebsd.org (8.8.5/8.8.5) with ESMTP id XAA19487
          for <freebsd-hackers@freebsd.org>; Sun, 20 Apr 1997 23:35:44 -0700 (PDT)
Received: (from bde@localhost) by godzilla.zeta.org.au (8.8.5/8.6.9) id QAA23521; Mon, 21 Apr 1997 16:31:12 +1000
Date: Mon, 21 Apr 1997 16:31:12 +1000
From: Bruce Evans <bde@zeta.org.au>
Message-Id: <199704210631.QAA23521@godzilla.zeta.org.au>
To: bde@freefall.freebsd.org, thorpej@nas.nasa.gov
Subject: Re: cvs commit: src/sys/i386/include endian.h
Cc: freebsd-hackers@freebsd.org
Sender: owner-hackers@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

> >   Don't pollute the namespace if _POSIX_SOURCE is defined.  This is
> >   broken - it makes almost everything in <machine/endian.h> inaccessible
> >   if _POSIX_SOURCE is defined, yet <machine/endian.h> is not a POSIX
> >   header.  Other systems don't do it any better.
>
>...there's a very good reason to protect the namespace:
>
>basalt (thorpej) ~ 202% uname -sm ; grep endian.h /sys/sys/types.h
>NetBSD hp300
>#include <machine/endian.h>

That's why the namspace protection implementation is broken.  After

	#define	_POSIX_SOURCE	1	/* protect namespace */
	#include <sys/types.h>		/* or anything that includes this */
	#undef	_POSIX_SOURCE		/* unprotect namespace */

the idempotency ifdef prevents further inclusions of <machine/endian.h,
so there is no way to get at the non-POSIX things in it.

It's interesting that the 4.4Lite byteorder(3) says to #include
<sys/param.h>, not <sys/types.h>, to get endian features.  This permits
a better implementation (except for <sys/param.h> being inherently
namespace-polluting).  NetBSD has fixed the man page.

Bruce