From mboxrd@z Thu Jan  1 00:00:00 1970
From: presotto@closedmind.org
To: 9fans@cse.psu.edu
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Message-Id: <20011126201949.C6129199E8@mail.cse.psu.edu>
Subject: [9fans] mine's smaller than yours
Date: Mon, 26 Nov 2001 15:19:48 -0500
Topicbox-Message-UUID: 2a87d64a-eaca-11e9-9e20-41e7f4b1d025

The last time I did a comparison was with Mach.  I tried
to compare the sources required to make a MIPS kernel with
nothing but a serial console as a device.  I just counted lines
with semicolons, i.e., 'grep ';' | wc -l'.  I really couldn't
configure Mach to do that without a Unix personality
so I didn't count that.  It was still an astounding difference,
on the order of 10 to 1.

If someone feels like doing it for Linux, Unix, whatever,
lets just come up with a benchmark configuration. How
about

x86
one vga card (nvidia tnt2)
ps/2 mouse
ethernet (intel 82557)
ip stack
keyboard
no disk, no local fs

Build the minimum kernel that lets you run enough cruft
to start a window system and get work done.  Assume a remote fs
(or not if you think that's unfair).

See how many ; lines that takes.

This doesn't mean that they have equal function.  However,
its the only objective comparison I can come up with.

What the results would imply is pretty subjective.  If
one group thinks memory is free, they might require all
sorts of stuff to exist in a minimal configuration.  Unless
the size of the compared systems is orders of magnitude
different, all you're really comparing is design choices
made by the designers for what they thought a typical
configuration would be.