Index Home About Blog
From: corbett@lupa.Sun.COM (Robert Corbett)
Newsgroups: comp.lang.fortran
Subject: Re: declaration amoung executables
Date: 28 May 2000 18:58:11 GMT

In article <8gr8m7$q93$1@sshuraac-i-1.production.compuserve.com>,
Michael Metcalf  <113547.3263@CompuServe.COM> wrote:
>When this feature was
>standardized, the committee, for whatever reasons posessed them,
>chose a
>slightly different syntax than what was already widespread.
>===========================================================
>
>Actually, Kurt Hirchert made a survey of all the exiting
>implementations at that time and discovered that they all
>differed in some way. In order not to favour one vendor, it was
>decided to adopt for f90, again, a somewhat different syntax.
>(This happened at the Bellevue meeting for which I appear to have
>lost my notes.)
>
>Regards,
>
>Mike Metcalf

I applaud your intent, but as an employee of a Fortran compiler
vendor, I must tell you you did not do us a favor.  Our customers
already expected us to be compatible with all our major competitors.
Now, we have to try to be compatible with all our major competitors
and with the new syntax from the standard.  We would have been better
off if you had chosen to standardize on the syntax of any of the
major vendors.

I implemented a strict implementation of namelist input as specified
in the standard in Sun f90 2.0.  That was a serious mistake.  Our new
release includes all the vendor extensions our customers told us were
missing that are compatible with the requirements of the standard.

					Sincerely,
					Bob Corbett


From: corbett@lupa.Sun.COM (Robert Corbett)
Newsgroups: comp.lang.fortran
Subject: Re: declaration amoung executables
Date: 29 May 2000 21:37:10 GMT

In article <8gtbq8$rd0$1@sshuraaa-i-1.production.compuserve.com>,
Michael Metcalf  <113547.3263@CompuServe.COM> wrote:
>
>I applaud your intent, but as an employee of a Fortran compiler
>vendor, I must tell you you did not do us a favor.
>============================================================
>
>Actually, it was more complicated. X3J3 decided by 15 to 9
>(including me!) with 6 abstentions to replace the then
>name-directed I/O facility by NAMELIST in 1986. I was mandated by
>CERN at the time to oppose the inclusion of NAMELIST in the
>standard as it was (is) an esthetic abomination. Name-directed
>I/o would have been the clean way to go.
>
>Regards,
>
>Mike Metcalf

Nonetheless, once X3J3 chose to include NAMELIST I/O, it should
have provided a specification for NAMELIST I/O that was close to
what was implemented by the major vendors.  That would have been
better for both users and implementors.

The Fortran committee has repeatedly invented new ways of providing
the functionality of widely used extensions rather than standardizing
existing practice.  I have repeatedly heard the rationale that the
committee does this to avoid favoring one vendor over another.  A
fact of life for Fortran vendors is that they must support any
language extension that is widely used.  Providing a new way of
doing something does not mean that vendors can stop supporting the
old ways, it just means they will have to support both the new and
the old ways.

						Sincerely,
						Bob Corbett

Index Home About Blog