Subject: Re: src/dist is a *bad* idea
To: Ken Hornstein <>
From: Greywolf <>
List: current-users
Date: 12/13/1999 23:09:08
Regarding Mr. Woods, from my own misguided point of view:

I note that if we can come down on an issue in a BSD way or a SysV
way, he *tends* toward SysV.

To his credit, I don't think it's just because it's SysV,
no matter how much some of us might do the opposite (myself
included) because it's *not* BSD.

Nonetheless, there are certain personal attachments to this thing
we call Berkeley Software Distribution.  Having had experience with
innards and outards of both System V and BSD, I will say this:

"And I thought SysV smelled bad...on the _outside_...".

Many people will agree with this, which is why the init.d/runlevel
thing as it stood/stands will rear its head, flame up and die, rear
its head, flame up, die, etc. ad nauseum.

Anyone else ever see the original SVR4 and how it mangled NIS so badly
it had to be glued on with epoxy?

Anyone ventured into NIS+?  Talk about ucking FUGLY!

System V did exactly these things right:

	echo:  The only thing better than sysv echo is printf(1).
	date:  A formattable date string without using awk.  That's nice.
	touch:  The ability to update files to a given date is cool.
		(It's also a potential security hole from an accountability
		 standpoint unless you do C2 mode-change accounting and
		 the like, but that was what made it kind of cool :-)
		[please do not expound on this into this OR a separate

How long did it take them to get the tty driver right?  (Never -- they
put Berkeley's into SVR4, mangled just a bit to get the SysV syntax)

But to be fair, System V has given Berkeley one thing that it needed
in order to keep going:


Its codebase sucked rocks, but it had the inroads.  This has lit the flame
to be carried, and I dare say that at times we tend to carry it on
our shoulder instead of in our hand.

Sorry to ramble -- I have this propensity for examining both sides of
the coin a lot.  That said, I don't agree with either the philosophy
or the design of init.d/rc*.d or runlevels.  I can not think of times
where I'm going to be flipping back and forth between configurations
so fast that I require a separate runlevel to define them.

Of course, I think I'm among the crowd that loaded the tu8 (?) boot
tapelet on the console, formatted and labeled in standalone mode,
loaded the miniroot from a 1/2" 600bpi tape, followed the instructions
to the letter, edited the configuration files by hand (on a hard-copy
console!) and had a running system about three hours later. ISO-mongers, FTP-suckers and HD-transfer jockeys have got it
SO easy by comparison! :-)

[I also ramble after not enough sleep, so I'll go nighty about now.]

NetBSD: Why use Gates and Windows when you can have doorways and tunnels?