Subject: Re: Proposed rc.d changes....
To: None <>
From: Ken Hornstein <>
List: tech-userlevel
Date: 05/06/2000 01:26:51
># >From my point of view, a new startup system was proposed.  Various
># people made comments about it.  I fully believe that the general
># consensus was that the new system was a good thing.
>What are you smoking?  I haven't seen a general consensus stating anything
>to that direction.

Sigh.  I can only disagree with your interpretation.  This has been
an ongoing discussion for YEARS.  The general agreement (and if
anyone ELSE disagrees with me, please come forward) has been that
rc.d stuff is fine, and that runlevels are not necessary (I don't
want to rehash the arguments for/against rc.d and runlevels, but
I strongly believe this is a belief held by the majority of the
NetBSD community; and no, I can't easily prove it).  I have seen
objections to this particular implementation, but I view that as
implementation differences, not differences with the whole rc.d

In fact, in my mind the use of rc.d was decided a long time ago (years).
We're just now getting to the implementation.

>Granted considering they came from roughly the same direction each time.
>But the thing is that those of us who do dissent have been told, more or less,
>that our opinions don't matter.

Well, that's not my interpretation.  The messages _I_ saw were people
disagreeing with your position.  I, personally, have not found any of
the arguments you or any of the rc.d dissenters have made to be that

>I have seen _more_ _different_ _people_ sending in a disagreement of the
>proposed direction than I have seen agreement!  Where does this constitute
>that more people agree?  Is the fact that most people have remained silent
>on the issue a demonstration of consent?  There's more dissent here than
>agreement, and there's more apathy, it seems, than either.  But silence
>does NOT equal consent, no matter how you look at it.

Please don't confuse "apathy" with "satisified with the status quo".
I haven't spoke up before now because this had already been thrashed
out a while ago.  I didn't see the need to voice my opinion when plenty
of other people had done so in a much better way, _and_ it was clear
to me that the majority opinion agreed with me.  If I felt like I
was in the minority, I'd speak up.

It's hard to say what the average person thinks about this thread.
I can only _assume_ (and I am clearly labelling this as an assumption
which I cannot prove) that like me, they're rather tired of it.  But
they don't want to get in an endless argument, so they quickly hit
their delete key when messages come in on this topic, because they're
satisfied with the outcome (or perhaps not upset enough with the outcome
to object).

>The thing that I find insulting - again - is that no matter what arguments
>are brought up, no matter how well thought out they are, no matter how well
>based in working fact, the dissenters are being told that the arguments are
>not valid.

Well, here we have a problem.  I personally didn't find any of your
arguments compelling (I speak as a former system adminstrator who
grew up using a monolithic rc.local).  Now, we _could_ continue to
send tons and tons of messages saying the same things over and
over.  I think everyone is tired of that.  Now, since we have two
viewpoints who apparantly cannot be reconciled ... exactly what should
we do?  I think the most reasonable course was chosen - pick what the
majority has chosen.

>It's more than religious, it's technical.  Editing a splintering of files
>is a PAIN IN THE GOD DAMNED ASS compared to being able to edit one file.

And you wonder why I labelled this religious? :-)

>Or hasn't that sunk in yet?

Well, frankly, "no".  I find the rc.d system much simpler; not just
for 3rd party add-ons, but in every way.  I don't want to rehash these
arguments - just pretend I've inserted one of the other set of arguments
here in this thread.

>And I frankly find it rather insulting that a matter of technical merit
>is being dismissed as religious.

I'm not convinced that this is merit; this derives from things like
"personal experience" ... things I think which are hard to quantify.
I can only say that my experiences aren't yours, and while I have
my differences with System V-based OSs, the way system startup
works isn't one of them.

>Some folks who are here have migrated from running that POS called IRIX.
>Some of us have had to administrate other OSs whose configurations have been
>a nightmare, and up to the rc.d split, that nightmare had not managed to
>manifest itself inside NetBSD.  And now it's here, quite a few of us object,
>and we are not being acknowledged in the least.  Who are you, collectively,
>that you can just turn around and piss on your user community like that?

Okay, let's get some hard numbers here.

I went back over tech-userlevel and current-users in the Feburary/March
time frame, and decided to count the pro/con rc.d people.  I've taken
everyone's email on the subject and either placed you on a "pro" or
a "con" list, depending on my interpretation of your email.  If I've placed
you on the wrong list, please let me know (note: I decided to put people
on this list if I felt they liked rc.d in general, not our specific
implementation).  Also, if you're not on here at all and you want to be,
just pipe up (I've only placed people on this list who have said recently
on public mailing lists for or against rc.d.  And I left myself off).


Aidan Cully
Andrew Brown
Bernd Ernest
Bill Sommerfeld
Brett Lymn
Christos Zoulas
Darren Reed
David Brownlee
Eduardo Horvath
Feico Dillema
Frank van der Linden
Giles Lean
Greg A. Woods
Havard Eidnes
Hubert Feyrer
Jared D. McNeill
Jaromir Dolecek
Jason R Thorpe
Jay Maynard
Kevin P. Neal
Laine Stump
Luke Mewburn
Manuel Bouyer
Martin Husemann
Mason Loring Bliss
Michael Richardson
Nathan J. Williams
Perry E. Metzger
Robert Elz
Scott Aaron Bamford
Simon J. Gerraty
Takahiro Kambe
Todd Whitesel
Urban Boquist


der Mouse
John Nemeth
Secret Asian Man
Wolfgang Rupprecht
Soren S. Jorvang
Carrie (Bowen) Jones

So by my count, that's 35 people who want some form of rc.d, and 7 who
would not like an rc.d.  Let's assume I've gotten at least 5 people on
the wrong list ... that's 30 to 12.  Still almost 3 to 1.  I would
call that a consensus.

>In the past, objections were at least addressed.  If you can't address the
>objections, or choose to ignore them, you're no better than the software
>giants from whom we sought herein to take refuge.

Well, what exactly _should_ be done?  We have what in my mind was
a clear consenus that we should adopt some variant of the System
V startup scripts.  The objections that were raised were on the
order of "monolithic rc is the way to go".  How is someone suppose
to compromise in the face of such objections?

Now, I understand some people have objections to the current rc.d
scheme.  I personally have my problems with it (for the record,
I would have preferred a straight port of the Sys V S*/K* scripts,
no runlevels), but as I understand it, the particulars of how rc.d
has been implemented haven't been fully resolved yet.  If _that_
is your complaint, then I think that's still an ongoing process, and
I'd wait to see what happens.