Help us defend against VMS!

[Barry Shein]

Re: VMS fortran as "industry standard", reply to Larry McVoy et al.

I think you miss my point (I don't particularly disagree with yours.)

I claim Unix is the premiere scientific computing base, including
fortran, because Unix runs on the machines scientists need while VMS
doesn't. For example, high-end graphical workstations, Crays, ETA's
(well, almost) etc.

I can't imagine in this day and age that anyone needing
compute-intensive cycles and/or data visualization looks towards VMS,
1..6MIPs on the whole architecture range just doesn't cut it anymore.

I have no doubt that a lot of people who like fortran like VMS's
fortran-like language (after all, you can extend the language just so
much before you are leading your users down a garden path as they
unwittingly use these "nice" features and lose all portability to
other machines they may have access to.)

Now, it's perfectly possible to write portable fortran code under VMS,
I believe there's even some switch on the FORTRAN command to help
check your code for non-standard statements. But of course, you're
then back to the fortran language that's available on the other
systems, so much for any advantages. And, in my experience, scientists
and engineers have practical troubles with this, they seem frustrated
that the VMS run-time library is not available for their code, let
alone nuances of syntax. It can be a frustrating trap, I suppose
one could argue that ignorance has its costs.

Anyhow, to clarify my point once and for all: VMS has a nice (so some
tell me) fortran devpt environment and a non-standard fortran which
helps give it that reputation (and, I hear, a nice debugger.) My
claims about Unix being superior for compute-intensive science was
much more an observation that Unix runs on the full range of
architectural performance needed by these scientists to get what
they really want, results. Niceties are nice, but so are answers.

It's possible it comes down to that I was talking about production and
you're referring to devpt, although I claim that devpt is good enough
under Unix (the differences are really not that huge) thus production
should outweigh these advantages.

	-Barry Shein, Boston University