[Info-vax] OT: news from the trenches (re: Solaris)

[lists at openmailbox.org]

On Thu, 12 Mar 2015 22:05:30 -0400
David Froble via Info-vax <info-vax at rbnsn.com> wrote:

> >>> Yeah but you seem to have suggested there's a question whether VMS
> >>> can or should run on cheap hardware given the VMS customer is
> >>> typically a high end customer.
> >> You're suggesting that x86-64 isn't high-end
> > 
> > That's right!

How anybody could think Intel is high-end is beyond belief. Are you
conflating fast with good?

> Ok, don't just say it, defend that statement.  What about the current 
> x86 processors isn't up to it, whatever "it" is?

I have already pointed out several objections and I wasn't the only one to
have done so. I see you're only objecting to what I have said, and not to
what other people have said that expresses the same view. If you're really
interested in what I have to say (doubtful!) then go back and read what I
wrote.

If you want to argue I have other things to do. And I'm probably not very
inclined to write a treatise on Intel's track record of failures and
stumbles as it is well enough known, still affects Intel users today, and
is not worth rehashing when I have more interesting, non-combative ways to
spend my time.

> Speed?  With the exception of possibly (most likely) Power, nothing is 
> as fast.

Well, if you want to argue, as you seem wont to do I would suggest you at
least argue with the person you are picking the fight with. Are you
violently agreeing with what I wrote, perhaps accidentally?

I have already acknowledged the Intel is fast. In fact, Intel is not the
fastest, POWER is faster as you suggested, and the mainframe (probably not
fair to compare it, but since Hoffman's raising venture capital to port
z/OS to commodity Intel crapware let's mention it anyway) IBM has had
clocks over 5.5 GHz in production for a couple of years already. IBM
usually doesn't compare clocks because they're not a good basis for
comparison but as long as we're here now you know.

> Memory addressing (size)?  64 bit addresses seems to be overkill.

Everybody and his brother has 64 bit addressing. Intel wasn't the first, as
you might know, and they screwed it up, and couldn't get it done so they
used AMD's design, which is also broken. Badly! Surprise, surprise...

> Regardless of whatever their previous source, today's x86 CPUs is about 
> as good as it gets, today.  Shitty beginnings or not, run the benchmarks 
> and weep ....

If all you care about is speed for the dollar then the argument is valid.
But nobody said Intel wasn't fast. They're not the fastest but they're the
fastest for the money. Ok, maybe that means something to people who run
Windows server because it can't get out of its own way. Solaris runs plenty
good enough on SPARC. I'll take Solaris SPARC over Intel Linux any day of
the year.

Do you really want an ugly woman with a half million bucks of plastic
surgery? If you look too close you can still see all the scars. That's
Intel. Lipstick on a pig. And not just one layer and not just one color.

Don't get confused and think a CPU's goodness is based on how fast it is.
That is important but there are a lot of other things that matter more. If
you just write application code it doesn't matter just as you don't have to
be confronted with how bad your Toyota is if you a) never tear it apart and
see how shoddily it's made or b) don't have any basis for comparison except
with other underpowered, poorly-made cars that you also didn't work on. I
understand there are a lot of people in that category. But they shouldn't
argue about stuff they don't have any connection to.

> We have an application that is in use by multiple customers.  It 
> basically runs their business, wholesale distribution.  It has over 40 
> years of development behind it.  Lots of business logic there.
> 
> The application is mostly written in DEC / COmpaq / HP Basic.  We looked 
> at porting.  It would not work.  Probably easier to start over. 
> However, the cost was deemed prohibitive.  Probably cost less to pay for 
> the VMS port to x86.  Not that that much money would ever be available. 
>   So, no port, and no re-write.

This is what I have been saying. In many cases, porting is not viable. But
if you can't port application code in an HLL then do you imagine you can
port OS and applications written to a specific architecture (not just an
ISA btw) in assembler?

> I'm sure it could be.  I have no love for Intel.

Yes you do, you just said it's the best and you got red-faced when people
suggested otherwise.

> But, what else reasonable is there?

You've asked, we've answered, you don't seem to read the posts.

> It's not the CPU that's important.  It's the environment, and the 
> applications, and for quite a few VMS users, there is not a viable 
> alternative.  And thus, an opportunity for VSI.  And salvation for many 
> of us.

It depends who's saying that and it's not something that can be said in
isolation. Sometimes it is the architecture that makes the environment
possible. VMS was a premium product running on bespoke hardware. If it
moves towards a commodity crapware platform competing with commodity
crapware OS then things will probably end badly.

-- 
Please DO NOT COPY ME on mailing list replies. I read the mailing list.
RSA 4096 fingerprint 7940 3F02 16D3 AFEE F2F8  ACAA 557C 4B36 98E4 4D49