[Info-vax] New CEO of VMS Software

[Dan Cross]

In article <unevlq$16jak$1 at dont-email.me>, chrisq  <devzero at nospam.com> wrote:
>On 1/7/24 17:22, Dan Cross wrote:
> >[snip]
> > I remember hating it.  Coming from a more "traditional" Unix
> > background, it was ... weird.  Printing, storage management,
> > man pages, the security infrastructure, all felt gratuitously
> > different for no real reason.  You were almost forced to use
> > their menu-driven management tools, but as the USENIX button at
> > the time said, "SMIT happens."  It all felt very big-M
> > "Mainframe" inspired.  The compilers were very good, and the
> > machines were fast, but the developer tools weren't bundled and
> > I remembered fighting a lot of third-party software to get it to
> > compile and run properly.
> >
> > That was all weird because, on the 6150 ("RT") machines they had
> > offered a very nice version of 4.3BSD Tahoe plus NFS to the
> > academic community; clearly, people at IBM knew how to "do" Unix
> > right.
> >
> > Weirdest for me was the lack of a real console.  There was a
> > 3-digit 7-segment LED display that would cycle through various
> > numbers as the system booted up; things that would have been
> > emitted to a serial port on a VAX (or even a Sun) were instead
> > represented by random collections of digits, and there was a
> > book you had to look at to see what was going on if something
> > hung.  Something like "371" was "fsck failed on /usr."  (I don't
> > recall if that was the exact code).  Then were was a the damned
> > key, where the system wouldn't boot if it was in the "locked"
> > position.  Which sucked if the machine crashed for some random
> > reason.  I walked into a lab one day and the entire network was
> > down because all the machines had crashed over some network
> > hiccup and the damned sysadmin had turned everything to "Locked"
> > for some obscure reason ("it's more secure.")  I guess he was
> > right: it's certainly more "secure" if no one can use the
> > computers.  :-/
>
>The RS6000k here (7043/150) has a bios console, updated from
>ibmfiles.com.  Functionality as one would expect. including
>extensive diags.

Yeah, they fixed a lot of that once they went to PowerPC.  The
earlier POWER (pre-PPC power) machines had a problem driving a
framebuffer in early boot, hence the 7-seg displays.

>Yes, there is a seven segment display showing
>post and boot progress, but running headless, that could be a
>real advantage.

I can see the utility.  A UART would be better.

>Can't be sure about C compiler, but think there
>is one. Package management seems good, so just a few minutes task
>to install Gnu tools. Also, the file system layout is more or
>less as expected. Perhaps the early machines were as you describe,
>but not the one here.

Yeah, that's all post the time I was using them.

We customized them more than I honestly felt comfortable to make
them manageable back in the day; I ported the lpd suite from
4.4BSD to make printing more familiar, for example, and ignored
their mainframe-esque batch stuff entirely.

>You don't have to use the automated tools,
>smit etc either, but they do have their uses. Pretty cool, fully
>sorted system, in fact. Slightly different in some ways, but easy
>to get to grips with and find way around.

That's another reason I didn't like it; it's true that you did
not _have_ to use smit, but especially for mucking around with
filesystems, it was a lot easier.  But that was something of an
issue: we had heterogeneous networks with Sun, IBM, DEC, SGI,
HP, DG, etc, gear; the AIX stuff was just enough different to
make it really irritating to keep things in sync with the rest
of the network; all of the infrastructure we'd built up since
the time of running BSD on VAXen had to be modified to special
case the IBM gear.

> > Ha, yeah, my SPARC hardware down in the basement hasn't been
> > turned on in years: it's too expensive to run.
>
>Yes, archive server only now, powered up as needed, but at 600+
>watts with the drive arrays, before even pressing a key, totally
>unworkable :-). Still have SS20 and more to play with though.

Yeah, exactly.  Power and heat TDP just doesn't make sense in
2023.

> > The trend had already started before that, I'm afraid.  A lot of
> > former Sun people I know acknowledged that they tried to stick
> > with SPARC as a differentiator way longer than they should have,
> > and that they should have embraced x86 much earlier than they
> > did.  They had a head-start with the Roadrunner, but they gave
> > up.  Had they stayed with it, perhaps life would have been
> > different.
>
>Sparc is still quite competitive technically, if you look at
>the specs.

This I actually disagree with.  Niagra seemed like it would be
competitive, and it was on paper, but realized workloads didn't
man out.  The fact is, the architecture just got kind of maxed
out in the same way Alpha did.  It's sad, but here we are.

>It's just that Oracle have given up on it. Solaris
>always was a very secure and robust OS and

Ehh....  I've seen the source code.  :-)

In many ways, it's better than contemporary systems, but in
many other ways, worse.  It certainly has a more coherent feel
about it than, say, Linux; but in so many ways its behind the
curve.

>there are real
>advantages from running a non X86 architecture, from a
>security pov.

I kind of buy this in the sense that the "POP SS" bug doesn't
affect you when you don't have an stack segment selector to pop,
but on the other hand, things like speculation vulnerabilities
affect more than just x86, and Sun gear had its fair share of
security bugs back in the day.  I'm positive that there are
still others lurking in the code, it's just that hardly anyone
cares anymore to go and find them.

>I remember seeing one of the early Sun X86 boxes, a 386i,
>1999 ish. An awful machine, slow, expensive and underwhelming,
>even compared to Sun 3.

Surely you mean 1989?

>Everyone hated it, but what they did produce were the X86 PC
>on a card products, to plug in to sbus and pci Sparc machines.
>Would run almost as a standalone pc, with all the sockets on
>the card cage bracket, including vga video, keyboard, mouse,
>soundcard etc, but were also  highly integrated into the file
>system and desktop at the Sunos / Solaris side. Quite
>reasonable  performance for the time and remember running
>Lotus 123 and a raft of pc apps via one of the cards.

I remember those things.  Very cool, but sadly not enough to
stop the MSFT juggernaut.  They really should have gone with
McVoy's proposal to open source SunOS and rally the Unix
vendors around that.  But those sweet, sweet margins were just
too good to give up, I guess.

>Perhaps a bit hard on DEC, as one of the things I most liked
>about DEC was the integrity and attention to detail of the h/w
>and s/w. Ran an Alpha 500/400  machine for many years. Tru64 unix,
>a very solid OS. but bit by bit, became an orphan with little
>open source support and no real future pathway. In many ways,
>it's open source software that has made many platforms what they
>are today, and their success, or not. If i'm still annoyed at
>DEC, it's because they had some of the best product and minds in
>the business from a technical pov, but squandered the lot on a
>greedy and inflexible business model. Hubris is it's own reward
>etc.

I liked OSF/1 (OSF, of course, was known to stand for, "Oppose
Sun Forever").  Interestingly, OSF/1 was based on a microkernel.
:-D

I always felt like DEC kind of got cheated by adopting OSF/1;
all the rest of the OSF participants had committed to switching
from their proprietary Unix to OSF/1, but only DEC actually did
it (granted, they were coming from Ultrix, so...).  It has
always reminded me of those movies where the commander says,
"all of you who will volunteer for this suicide mission, step
forward" and then turns around, while the rest of the
assembled troops take a step back, leaving just the poor
schlep who didn't realized what was going on standing where he
had been, leaving him to be chosen.

>The really oddball unix ime, was HP-UX, where nothing is where
>one would expect to find it, and a whole shedload of oddly named
>commsnds, like learning a new language.

Ah, Hockey Pux!  Yeah, another oddball.

>Have a good new year anyway. Still some progress, with arm and
>Risc 5 likely to further upset the established order :-)...

The industry is rapidly shifting to ARM, and I hope we'll see a
competitive server-grade RISC-V offering in the next five years.

	- Dan C.