by homeworld.cygnus.com with SMTP; 27 Nov 1998 22:35:18 -0000
	by smtp04.primenet.com (8.8.8/8.8.8) id PAA22982;
	Fri, 27 Nov 1998 15:33:54 -0700 (MST)
 via SMTP by smtp04.primenet.com, id smtpd022977; Fri Nov 27 15:33:52 1998
	by usr02.primenet.com (8.8.5/8.8.5) id PAA20989;
	Fri, 27 Nov 1998 15:32:31 -0700 (MST)
From: Terry Lambert <tlambert@primenet.com>
Message-Id: <199811272232.PAA20989@usr02.primenet.com>
Subject: Re: Merging Net/Free/Open-BSD together against Linux
To: louis@signalpath.on.ca
Date: Fri, 27 Nov 1998 22:32:31 +0000 (GMT)
Cc: advocacy@openbsd.org, netbsd-advocacy@NetBSD.ORG,
        FreeBSD-advocacy@FreeBSD.ORG
In-Reply-To: <Pine.BSF.4.03.9811261320510.11270-100000@tronix.signalpath.on.ca> from "Louis Bertrand" at Nov 26, 98 02:11:04 pm
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit

> In my experience in the hardware domain, standards favour widespread
> adoption but stifle innovation.

You mean like FTP, SMTP, HTTP, HTMP, and MIME "stifle innovation"?

Or do you mean like ELF, DWARF, NROFF, and SGML "stifle innovation"?

Or perhaps ANSI, POSIX, and XPG/4?

I'd agree if you wanted to say "Bad standards, like ISA, stifle
innovation", though...

I think that software standards tend to codify interfaces, and that
hardware standards tend to codify implementations.

Very different spaces.

Not that I would mind rewriting all of libc to get rid off all static
or per thread allocated buffers, and to make all functions return only
status codes, forcing data returns to be implemented by passing a
return area by reference, mind you.  I would *love* to see:

	int
	ftell( FILE *stream, off_t*result)

	typedef u_int64_t	off_t;

(one example whre return values are overloaded to return error
information, to the detriment of the long term utility of the
interfaces).

I would also *love* to see a method whereby the argument sizes are
passed as part of the information so that system call interfaces
can be easily and safely versioned without proliferating entry
points.

But of course, that would constitute a "calling _standard_".


					Terry Lambert
					terry@lambert.org
---
Any opinions in this posting are my own and not those of my present
or previous employers.