On Fri, 31 Oct 2025 09:09:10 -0400, Phil Smith III <[email protected]> wrote:

Hi Phil - you inspired me to argue with myself. Or play Devil's Advocate - 
whatever.

> Yeah, I think the ship has sailed. People believe that cloud/clusters/whatever
> of Linux machines are Good Enough, and Good Enough is Good Enough. I haz a 
> sad.

So this is a good point that I omitted before, because it had slipped
my mind.

The theoretical scenario was a startup. So - are there circumstances
where a startup would choose IBM?

Yes, there is the single supplier - nominally.

However, first I would argue that that's the only supplier that isn't
a clown show. IBM has a business model that ensures that costs
are amortized and all problems will be solved. Everyone else is
on shaky ground, taking their chances with "free" software. I think
a lot of problems in the IT industry stem from anyone believing
that you can get something for free.

And yes - I'm one of those people producing nominally free
software myself - but I'm not expecting anyone to put their
production system on it. I wouldn't. It's missing the
"commercialization" step before you can think of doing that.

But let's say there is a sensible startup, that knows they don't
want to take their chances with clowns if they actually end up
being successful.

So there is literally only one choice for both software and
hardware. When they reach that point in time, is IBM's
charge going to be the biggest business expense they
have?

Even if it is - is that a problem? Everyone is in the same boat.
They just advertise themselves as a non-clown operation,
and pass the costs on to the customer. (You may run into a
problem here with customers themselves being clowns).

Ok, so if they theoretically see that as the end game, how do
they start out?

Just a suggestion - how about they start with a variation of
the clown show that doesn't rely on the expectation that they
can use 100% of the machine's capacity?

Specifically they run a variation of emulator - Hercules is the
most famous, but FSI has a commercial one too - including
z instructions (the ones they use should all be out of patent
now).

And MVS 3.8J can be run. Sure - you're restricted to 16 MiB -
but - so what? Perhaps that's part of the problem with the IT
industry - people expecting more than 16 MiB of memory.

Here's something someone else wrote:

https://www.quora.com/Why-did-you-leave-a-job-as-a-software-engineer/answer/Jeff-Sturm-2

What's the biggest genuine executable you've ever written?

The biggest actual executable with what I consider to be
"genuine" code is gcc 3.2.3. It is 400,000 lines of C code,
that translates into 700,000 lines of assembler code, and
produces a 3 MB executable. Easily fits within 16 MiB.

Is your startup going to produce a 400,000 line application?

After how many decades?

Can you even maintain that? What if you lose the one guy
who can actually maintain that?

By the time you get anywhere near that, I would expect you
to have moved to genuine IBM hardware.

My own code - written over 3 decades - and not completely
alone - comes to something like 70,000 lines of code, and
fits on a 360k floppy as main executables.

People quite literally ran MVS 3.8J for their production systems
too.

And then there can be other mitigating factors. If you write
your application in C90, you can at least theoretically move
to a non-IBM system if they did screw you over too badly.
C90 is just an example I am familiar with. Presumably a
"better" language could achieve the same result. There is
no legacy as you choose the language for your startup.

Be cautious wherever you take your startup, and keep an
eye on the fire exit at all times.

And you should be able to hop off z/OS and onto MVS 3.8J
if the worst happens. That can be your DR system. Even if
you have to have a reserve budget for a very powerful non-IBM
system (or maybe the cloud can rescue you there?) to run under
emulation, as you are forced to join the clown show due to
unforeseen circumstances.

And again - even if your main business system is running on
the non-clown IBM system, you can still offload less important
stuff onto clown systems. Not just MVS 3.8J, but the EBCDIC
x64 (ucx64e.zip from pdos.org) could have a role. Or a variation
of that. There is no need for it to be called "mini-Windows". It
could be called "mini-OS/2" or "mini-Linux" (ie those APIs can
be supported - to a similar functional extent - if required).

> And with the erosion of skills available, going TO IBM Z
> at this point would be very, very difficult and very, very
> expensive. I just don't see it, try as I might.

And this shouldn't be an issue either. A startup could be
up and running on MVS 3.8J pretty quickly. ie getting their
first "hello world" to run successfully.

And remember - this is all based on the assumption that a
startup has the attitude that they want to end up on the
mainframe like the rest of the Fortune 500. Aiming for
professionalism or whatever.

And again - just posing this as a theoretical possibility.

Otherwise the alternative is what? No new players will
ever be onboarded? Or maybe the original suggestion
was indeed technically correct - a $10k system will
exactly provide the necessary stepping stone? And IBM
so far hasn't seen the need for such a stepping stone
because they are legitimately short-sighted? Maybe IBM
could allow zPDT for this purpose as well - but again -
too short-sighted?

Or perhaps MVS 3.8J (or other alternatives) could force
IBM to make zPDT available for this purpose. But that
also self-evidently didn't happen. But maybe that's because
MVS 3.8J is missing one crucial component, like GNU
Cobol, that would put it over the edge?

What would you personally do if you were going to start
a startup? Given your philosophy expressed above, where
you personally can see the difference between IBM and
everyone else. Starting with - what language would you
even use? Starting a new bank, perhaps? How many
transactions per second does your new bank need to
be able to process? Can that be done under MVS 3.8J?

(Note - I personally have travelled a quite narrow path
through the IT industry - I'm not claiming to have any
answers - nor a broad understanding of anything in
particular - I just demonstrated some stuff physically
working - nothing more, nothing less - and not for any
particular purpose - I'm not expecting people to
manufacture physical EBCDIC ANSI X3.64 terminals
either - I just wanted to prove that they could be
created and used - I haven't proven they can work
under TSO on MVS 3.8J though - or perhaps - a dual
3270/EBCDIC ANSI terminal).

BFN. Paul.

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to