*BSD News Article 32946


Return to BSD News archive

Xref: sserve comp.unix.bsd:14333 comp.os.386bsd.misc:2688
Path: sserve!newshost.anu.edu.au!harbinger.cc.monash.edu.au!msuinfo!agate!howland.reston.ans.net!vixen.cso.uiuc.edu!newsrelay.iastate.edu!news.iastate.edu!ponderous.cc.iastate.edu!michaelv
From: michaelv@iastate.edu (Michael L. VanLoon)
Newsgroups: comp.unix.bsd,comp.os.386bsd.misc
Subject: Re: 4.4-lite?
Followup-To: comp.os.386bsd.misc
Date: 17 Jul 94 07:24:59 GMT
Organization: Iowa State University, Ames, Iowa
Lines: 95
Message-ID: <michaelv.774429899@ponderous.cc.iastate.edu>
References: <2vgvc7$3tg@spruce.cic.net> <301rrc$cmv@masala.cc.uh.edu> <Csxxqo.DsM@newsserver.aggregate.com> <3097eh$m2h@pdq.coe.montana.edu> <Bs2yi5F.dysonj@delphi.com>
NNTP-Posting-Host: ponderous.cc.iastate.edu

In <Bs2yi5F.dysonj@delphi.com> John Dyson <dysonj@delphi.com> writes:

>Nate Williams <nate@bsd.coe.montana.edu> writes:

>>Unless the port is to a completely different platfrom.  The move to 4.4
>>has placed both groups on level ground for the most part, so getting
>>another BRAND-NEW (PPC, ALPHA, etc..) platform going is going to take
>>the same amount of work for both folks.

>And remember -- the FreeBSD VM kernel enhancements are already in our
>2.0 working sources.
[...]
>So, yes when
>it comes to porting platforms -- FreeBSD and NetBSD SHOULD be on nearly
>level playing grounds -- but the FreeBSD kernel, when it comes out will
>be as advanced as ever.

John, I don't mean to be hostile, so please don't take it that way,
but several things you're saying just don't quite seem to sync with
me.  I'm sure you have the best of intentions, so please bear with me,
and enlighten me if you can.

Nate and you say that both platforms are on even ground when it comes
to starting new architecutre ports from scratch.  Yet, the NetBSD code
already has all the hard to find endien problems all fixed, and is
withing weeks of "shipping" a 4.4-based system, NetBSD-1.0, that is
releasing on more architectures than can be counted on one hand.  It's
already done there!  The compat subsystem is mature and fully
debugged; the different arch layers are well defined; a good system
for implementing all the different parts is in place.

The NetBSD team has spent months converting to an architecture neutral
system, and sorting out all the hidden problems that come with it.
And even in the conversion to 4.4 stuff, we still saw little endien
related bugs pop up that had been buried deep, and had to be fixed.
Active development on other platforms made identification and
elimination of these kinds of bugs infinitely easier than it would
have if working from only one platform and trying to code towards the
general case in the dark, so to speak.

Now, every time somebody mentions that NetBSD has a lead in the
multi-platform department, you say FreeBSD is right there behind them.
It seems to me that you severely underestimate the amount of work it
will take to get your code truly architecture neutral from top to
bottom, and to get another architecture, especially a different endien
architecture, running as completely as the iX86 one.

Am I missing something?  Has the FreeBSD team been doing some
undercover architecture work that I haven't heard about?  I'm serious.
I haven't seen any outward signs that you have, so please give us the
details so the guessing can stop.  I don't think you give the NetBSD
guys enough credit for the hard work and pain they've gone through to
make the multi-arch support in NetBSD work as well as it does.  And,
unless you guys have put in a *lot* of hours on the same kind of
stuff, it makes me feel like you haven't taken a serious look at what
needs to be done, yet, when you say FreeBSD is just as ready.

Understand, that I didn't take part in much of the NetBSD code
development, but I have been a user of NetBSD-current for quite a long
time, and I got to watch the development process in action while much
of this was going on.  And I can tell you that there were some weeks
when you wondered if it was even worth it.  But the end results prove
that it definitely was.  In short, there were some very difficult
times in the transitional areas.

I'm not a core member, and I don't speak for anyone on the NetBSD
development team.  And I don't want to slam FreeBSD -- it seems to me
like you guys have assembled a great OS, and have a great following.
And I've dealt privately with various members of the FreeBSD team at
times, and they all seemed like likable guys.  But I want to clarify
this issue because, it seems to me, you take away from the legitimacy
of NetBSD when you claim that FreeBSD is just as capable in the
multi-arch department as NetBSD, when I know what kind of pain NetBSD
went through to make all that stuff work.  And I see this claim every
time someone posts about how NetBSD is the system to chose for
multi-platform development.

I know you feel your OS is the best there is -- your own work is part
of it.  That's entirely understandable -- I would feel the same way.
But it seems to me that you always tell everyone that yours is the
most advanced and can do everything as well as NetBSD, when we all
know NetBSD is every bit as advanced as FreeBSD, only in different
areas.  NetBSD does some things better than FreeBSD, this being one of
them; FreeBSD does some things better than NetBSD, like maybe the
installation process and your bounce-buffers for ISA-bus DMA in big
ISA machines.  Please try not to detract from NetBSD in your (well
founded) enthusiasm for FreeBSD.  As a use of NetBSD-1.0-Alpha, I can
attest to it being one helluva system as well, and it would be nice if
you'd acknowledge that a bit more when you make claims about the
abilities of FreeBSD.  They're both excellent OS's.  And if you have
some previously unknown developments in the architecture department
under FreeBSD, please fill us in.

Note: I've redirected followups to comp.os.386bsd.misc.