4.3-VAX distributions

[Steven M. Schultz]

Hi -

> From: SHOPPA at trailing-edge.com
> To: PUPS at MINNIE.CS.ADFA.EDU.AU

	All caps?  Must be using a V(erbose)M(essage)S(system) confuser - didn't
	know there were any left :-) :-)

> Exactly.  The details are all in tmscpboot.c.  Prepend this to the
> "tape directory", write it to TK50, B MUA0, and you're at the "="

	Since PUPS is on a uVax kick at the moment I'll chime in with my
	(not so fond) memories of trying to jack 4.3-Reno onto a uVax-II.

	It was a perverse sort of fun but not something I'd willingly do
	again.  Burnout?  Perhaps.

	the base 4.3 system up to and including Tahoe couldn't be cold started
	on a KA630 (much less a 650 since that didn't exist yet ;)).  You _had_
	to have the Ultrix 'boot' bits&pieces to work with.  The 4.3 kernel
	had uVax support in it but the boot stuff did not.

	With 4.3-Reno that changed but...  As others have noticed the 
	cold start kit didn't create tapes suitable for a uVax.

> prompt, from which you can execute the standalone images.  "format"
> seems to crash badly, but one probably doesn't need that on a Q-bus
> machine :-).

	What I ended up doing was using my 2.11BSD 11/73 to create a bootable
	4.3-Reno tape for the uVax - all the pieces are there, just need a
	system to 'dd' the files out with the right blocking factors, usw.

	Then the fun really began.  The SPL "probing" logic in the kernel
	had a small problem when probing for MSCP controllers.  As I recall
	(and this is going back quite a few years) some 3rd party adaptors
	ran at a different (lower) SPL than the probing logic expected - thus
	the autoconfig routines raised the SPL higher than the interrupt
	of the (Dilog I think) controller and the whole system hung.    So,
	to install the system you HAD to use DEC controllers - ok, I had a
	RQDX3 and a couple RD53 drives present (the Dilog had a 319mb Miniscribe
	disk).  BUT 4.3-Reno had a bug in the MSCP driver and would not 
	recognize an honest to DEC RD53 drive!  This was rapidly getting to be 
	unfun.  I think the workaround (it's been a __long__ time so memory
	is fuzzy) was to lie and call the drive an RA60 and then correct the
	problem later.  But to get the lie thru to the kernel I had to 
	use the standalone 'copy' program to copy a file (created on a PDP-11)
	to the first couple sectors of the uVax's RD53.  Sheesh!

> The compiled-in partition tables used during an install are a real
> pain compared to, say, a 2.11BSD installation, where disklabel is
> a standalone utility!  (That's a real win, Steven!)

	You're quite welcome.  Actually 4.3-Reno served as inspiration and
	reminder of pain to avoid when it came time to implement 2.11BSD's 
	disklabel capabilities.  I swore I'd never go thru the pain of the
	kernel having labels but the standalone utilities lacking them

	4.3-Reno did have disklabels (the first 4.3BSD to do so) BUT the 
	standalone programs still had compiled in partitions.

> >First, I'm using Ultrix as my cross-compilation base, not 4.3BSD-Reno. (I
> >would say there is less of a gap between 4.3BSD-Tahoe and Ultrix than
> >between Tahoe and Reno. The latter is really huge, it's a gap between True

	Can't be any version of Ultrix I ever used.  At the time 4.3-Reno
	came out Ultrix was still a warmed over 4.2BSD that DEC had corrupted 
	with System V(anilla) bolted on contamination.   Affectionately known
	as Buglix ;-)  That was the same era that DEC had Ultrix-11 and that
	was a mucked up 2.9BSD.  Of course you have to realize DEC had  "Mr.
	Ken (Unix is Snake Oil) Olsen" around at the time 8-)  UNIX is still
	around - but DEC?  No, I don't like Compaq confusers thank you ;)

	4.3-Reno was a transitional experiment that happened just as the CSRG
	and DEC had a serious falling out - and DEC support (Vaxen) vanished
	at that point.  Any further work (4.4BSD) totally and completely
	ignored all DEC machines.

> Gees, looking at the install docs there are some very real improvements
> in Reno, especially in the filesystem and the speed of recompiled

	Yep - you get NFS (which no 4BSD had prior to Reno).  NFS doubles
	the size of the kernel though (at least) so there's a memory penalty
	to pay.  It also brought many of the POSIX features (termios for
	example).

> code.  I'm willing to live with a bit more disk space usage, especially
> for the promised speed benefits.  It's not like KA630's or KA650's are speed
> demons, and big cheap disks are readily available these days.

	Disk is cheap.  Especially for older drives (but you run the risk
	that an old drive will die soon ;-().  Best to invest in a modern
	SCSI<->MSCP adaptor and use current drives (that's what I did for
	my 11/73 - adaptor is $$$ but the drives are cheap).

	Boy, you're not just whistling Dixie (apologies to those outside the
	US for which the reference is obscure).  "Not a speed demon" doesn't
	begin to describe it.  I went, believe it or not, thru the work of
	getting a newer GCC-2 (at the time I think 2.3.x was "new") to build
	and run on a uVax-II under 4.3-Reno.  The biggest problem was that
	4.3-Reno was neither "old" (V7ish) Unix or "POSIX" (just getting off
	the standard's writers desks).  Getting GCC to build was a stop/go
	effort for several days but in the end the build would work:  about
	23 hours (or so)!!   Sheesh - a 11/73 can *completely* regenerate 
	itself from sources (all programs, manpages, etc) in about 28 hours.

	It was an interesting experiment but the uVax-II has sat here for 2+
	years without being powered up.  At one time the thought was to port
	4.4BSD over but everyone that _could_ do the work lost interest - I've
	my PDP-11s and PPro systems to keep me busy so I haven't the time or
	inclination to do much with a KA630 system.  For "slow" I have a PDP-11
	(lots of fun, keeps you humble with the address space limits ;)).  For
	'fast' I have a couple dual cpu PPro systems (running BSD/OS) that
	can give a quad processor SUN Enterprise Server-4500 a run for their
	money.  I have no need of a "slow" computer that attempts to run
	current day (bloated) software.

	I've toyed with the idea of swapping the innards of the 11/93 and
	the uVax.  The KDJ11E would be a lot happier in a BA-123 than a BA-23;)

	But that's as far as it's gone (thinking about it).  So - if anyone
	out there wants a uVax-II (9mb of memory but lots of disks and a 9-track
	tape drive to go with) drop by my place (shipping's out of the 
	question).  If you're more hardware capable than I perhaps we could
	swap the stuff into a BA23 (smaller enclosure to drive home, ...).

	Yikes and gadzooks - I was a bit verbose tonight (but my typing skills
	are much improved! ;-)).

	Steven Schultz
	sms at moe.2bsd.com

Received: (from major at localhost)
	by minnie.cs.adfa.edu.au (8.9.1/8.9.1) id UAA14865
	for pups-liszt; Sun, 22 Nov 1998 20:31:53 +1100 (EST)
	(envelope-from owner-pups at minnie.cs.adfa.oz.au)
X-Authentication-Warning: minnie.cs.adfa.edu.au: major set sender to owner-pups at minnie.cs.adfa.oz.au using -f