To: Ted Lemon <mellon@fugue.com>
Cc: Jonathan Stone <jonathan@dsg.stanford.edu>, perry@piermont.com,
        developers@NetBSD.ORG
Subject: Re: What is the limit on kernel size? 
In-Reply-To: Your message of "Wed, 13 Mar 1996 08:53:54 PST."
             <199603131653.IAA29854@toccata.fugue.com> 
Date: Wed, 13 Mar 1996 12:28:02 -0500
Message-Id: <12091.826738082@UX2.SP.CS.CMU.EDU>
From: Chris G Demetriou <Chris_G_Demetriou@ux2.sp.cs.cmu.edu>
Sender: owner-netbsd-developers@NetBSD.ORG
Precedence: first-class

This is my last message on this topic to this mailing list, period.

> If you run a kernel out of a compressed
> filesystem image, you have to have the bootblocks do the
> decompression, which according to pretty much everybody who's weighed
> in on the issue (and I would like to include myself here) is a major
> hassle.

I actually would disagree; putting decompression in a boot-block-like
program _would not_ be hard.


> I'm sorry, Chris, but for some reason I'm
> entirely missing the point of why you're so dead-set against this.

I think there are several issues here:
	(1) i have not been convinced that it's the 'right thing'
	    to do,
	(2) i have no confidence that the implementation that perry
	    would be adopting is in any way good, and,
	(3) frankly, i have very little confidence that perry would
	    make 'the right decision' for me (as a developer of
	    another port), and so i'm _not_ going to take his "OK"
	    on an implementation without question.


In my little world, (1) is largely dictated by what works, what can be
used in a nicely machine-independent way, and what is 'pretty' (not
necessarily in that order 8-)".

confidence for a change is gained by showing -- not ranting -- that
a change is the 'right thing.'

confidence in a person's abilities is gained by having the person in
question _do_ something, i.e. make good decisions and carry them
through to completion, rather than repeatedly saying "this is what i
would do," etc.  If you're not confident enough in your own decisions
to actually go an implement them, others be damned, just to try them,
then why should others be?


Perhaps that's insulting, but quite frankly, that's what I think my
objections boil down to.  If, say, Jason said "this is the right
thing, and this is the way it is going to be done," i probably
wouldn't argue with him much at all, because i'm confident in his
abilities as a 'reasonable' designer and programmer, because of the
previous work he's done.

I don't think it's wrong -- in fact, i think it's a _good_ thing -- to
demand that people who aren't "known good implementors" justify their
plans, design, etc., before those designs are accepted as "the right
thing."  In fact, i think even "known" good implementors should do
it...  It's the notion of peer review, i guess; it's why i try to do
implementations of things, and get them sanity checked by people
before i check them in, why i rewrite so much code i've already
written, etc.  It's why, say, the FreeBSD folks are very carefully
checking the changes they bring in from Lite2...


It's not even clear to me, from Perry's messages, that he understands
the design of the code (if there is a design 8-) that he'd be using,
and understands what alternate designs might look like.  Certainly,
from the start of this discussion he's said a lot of things about
"boot blocks" in general that are just aren't true on ports other than
the i386.  If he implements some cool install program for the i386,
i'd like to be able to use it on the alpha.  If i can't because it's
badly structured because he didn't bother knowing, asking, or even
thinking about how other ports would work, well, that's a screw, and
may not even be an 'incremental improvement' for any of the other
users of the source tree.


cgd