Subject: Re: Memory over-commit (Was to do with grep(1))
To: None <firstname.lastname@example.org>
From: Lucio De Re <email@example.com>
Date: 07/14/1999 12:34:11
On Wed, Jul 14, 1999 at 10:35:29AM +0200, Robert Elz wrote:
> I have no doubt but that you can dream up scenarios where you pander to
> the laziness of programmers, ...
I'd like to introduce a political note here:
It is by "pandering" to various aspects of human nature, be it laziness,
a zest for novelty, simplicity, excitement, that Microsoft and other clever
market-oriented organisations win the heart of the public.
I detest the effect this has on _my_ life, but I cannot deny the validity
of the approach. What Matt is repeatedly saying, is that there is nothing
like sufficient demand for FreeBSD to lead the way in providing robustness
(you may all find the information at <http://www.eros-os.org> interesting
in this matter, and then there is Plan 9 and Inferno for something in
between), and there is truth in what he's saying.
I've long suspected that my aversion to "everything-and-the-kitchen-sink"
syndrome of Linux and, to a much lesser extent FreeBSD is the reason for
sticking to NetBSD even though I always fall back on the above for client
installations (that's because the client can call on someone else for
support, which is improbable with NetNSD, not because I don't trust the
What I'm suggesting, is that NetBSD _is_, in practice, HobbyBSD, or, less
insultingly, ResearchBSD, and the UVM and other exciting developments
bear this out. That's how I like it, and I guess is what attracts most
everyone else out here.
On the other hand, that's not going to popularise NetBSD, so we'll always
be second-rate siblings: the nerd that never quite made it with the girls.
Me, I can live with that, in fact, somewhat cherish the role. Specially
as it saves me from attempting to evangelise, something that eventually
all other OS proponents seem to fall into. My understanding is that NetBSD
may excel where other OSs have little interest. If WinNT is a measure
to go by, this is a recipe for commercial failure, but it is also a quest
for perfection that commercialisation and time-to-market issues clearly
prohibit. I'd like NetBSD to stick to this perception, and I sincerely
hope I'm not mistaken in my view.
Going back to overcommitment, I believe that self-discipline is
the approach to be encouraged. Whether I turn over-committment on
or off, I'd like _my_ applications to be able to respond to critical
situations within the multiuser environment and attempt to contribute
in some way to alleviating the problem. If AIX is leading the way
in that direction, well, we should adopt their approach. Certainly,
the "kill -9" approach is insane, it has the logic (quite close to
the bone here in Africa) of "feeding the homeless to the hungry."