psmith@mozart.uucp (Presley Smith) (09/02/89)
In response to Keith Bierman, Kurt Hirchert, and Achille Petrilli...
This has become a religious war with neither side convincing the other
side of anything... it's time to summarize positions and move on to
more productive things...
First, I'd like to challenge EVERYONE, who is reading and digesting the
latest Fortran 8x draft that is out for public review, to review your
FORTRAN 77 standard WITH the 8x proposal and help find any places where
FORTRAN 77 might have been changed in Fortran 8x. If you find something,
send it in as part of your public review or post it or whatever...
Just a few more comments:
1. I, for one, am NOT convinced that FORTRAN 77 has not been changed
in some ways others than those documented in 1.4.1. Do I have
any other specific examples at this time, NO. With each review
of the document, I have passed all the information that I've
found to X3J3.
Several things make me worry:
- Many things have been voted on in the X3J3 committee with "text to be
supplied later..." or "person x knows what to change". I know of NO
other committee that operates that way. Every other committee that
I've been a member of has operated ONLY on proposed changes with line
numbers and text for a particular document. NO vote on adding
or removing text was taken prior to seeing proper documentation
of what was going to be changed and where the changes were going to
be made. X3J3 cannot produce a documented trail to show what was
changed for some of the changes that have been made.
One example of the problem of getting the Fortran 8x document
correct has been on this net. The issue of removal of "with the
same interpretation" in TWO places in the document WAS DEFINED
in a DOCUMENT voted on by the committee which contained BOTH
PAGE and LINE NUMBERS to be changed for BOTH places in the
8x draft, COMPLETE with the TEXT to be SUBSTITUTED in each case.
Due to human error or whatever, it was only changed in one
place. This seems like a trivial error. How many more serious
errors are there?
I also know that the differences in padding discussed in 1.4.1
was found by a person studying the document. Also, the committee
has NOT kept a list of changes that affect FORTRAN 77 compatibility.
1.4.1 was added to the current public review document between
the 1st public review and the 2nd public review because the SD-9
requires such things to be documented.
- This committee spent the last several meetings adding or re-adding
things to the proposed standard to attempt to address the public
review comments. It's these quick additions and changes that have
created what Walt termed the "messes" in certain areas of the proposed
standard. Integration of some of these features into this complex
document requires time and effort to insure that all the places in the
standard that are effected are changed properly. Proper aging of
the document to find and clean up these problems has not occurred
because X3J3 was under mandated deadlines.
- The chair of X3J3 has scheduled an extra meeting of the X3J3 group in
October to make additional corrections to the proposed standard
that YOU are now doing the public review on. The goal of this
extra meeting is to NOT add any functionality but to fix problems
in wording, etc. in the document is out for review. Errors that
are found by the public review will be addressed at the January
meeting of X3J3. Hopefully, many of them will already have been
corrected by the October X3J3 meeting.
- No one has tried in practice to do many of the things that are
contained in this standard. One example is a debate over a paper
presented at the Vienna meeting on a string module that had been
written in Fortran 8x. The problems encountered were discussed
and the author gave several suggestions for redesign of certain
8x constructs that he had tried to use in this module. Debate
had to be cut short due to time constraints...
- I know from experience that many times when we make major changes
in either software or documentation at the last minute to "fix"
some problem, we tend to break other things without realizing it.
And, it many times costs more than it should in both time and money
to get it all right again before it can be shipped.
In summary, I am NOT in any way implying or saying that X3J3 has
deliberately changed FORTRAN 77 in the proposed standard. I know
that so many changes have been made so fast that there have been many
opportunities to introduce errors and little time to discover and fix
those errors. My experience tells me to QUESTION this situation. If
your experience is different, you can certainly ignore my comments.
Kurt is a member of X3J3 and has been for many years. I've been on
X3J3 for 2+ years and on 3 other ANSI committees before X3J3. I
understand the process of standardization and have been a part of
many debates on X3J3. Kurt has also watched what has happened in
the meetings of this committee in the last two years... and he
can certainly correct anything that he believes is wrong about what
I have just said.
2, The second point is that there's too much protest over keeping
keeping FORTRAN 77 as an active standard. Everyone is out to
defend their position. Achille and Keith want to take every
word I say apart and tell me there's NO problem! Anyone that
tells me "don't worry, everything is all right" makes me suspicious
immediately. I've sat in too many meetings on this standard, that
Keith and Achille have not been part of, to not have some worries.
People tend to react poorly when something is forced on them
when they don't believe it should be. What is not a problem
to Achille or Keith may be a problem to someone else. If it
was clear cut that everyone wanted Fortran 8x, then I would expect
to be bombarded with requirements for it from customers. I'm currently
not seeing much demand.
3. There appears to be a lot of confusion over the SPARC resolution.
The complete text of the resolution is:
SPARC recommends to X3 approval of the attached Project
Proposal for Project 67-D. SPARC noted that this revised
Project Proposal calls for retention of the current ANS X3.9-1978
and the processing of the new FORTRAN 8X proposal as a
separate standard. SPARC also noted that, in making this
recommendation, it is intended that the current ANS X3.9 not
be the subject of future revision, consistent with X3J3's
concerns.
The changes in the Project Proposal are to call Fortran 8x,
"FORTRAN Extended."
- SPARC specifically noted what X3J3's concerns and said that
FORTRAN 77 would NOT be the subject of future revision.
- This is only a United States issue anyway. This resolution
has to do with NUMBERING of standards in this country and
what standards are ACTIVE at any time in this country.
- This resolution did NOT ask ISO to take the same action.
ISO is free to do what it wants with FORTRAN 77 and Fortran 8x.
- FORTRAN 77 will have to be re-affirmed again in 4 years for it to
continue to be an active standard. If not, it will automatically
be dropped as an active standard. (For those of you that did not
know, FORTRAN 77 was re-affirmed last in October, 1988 so that
we continued to have a FORTRAN standard. Without quick action
on the part of X3 and X3J3, there would be NO FORTRAN standard
in the United States today. It will have to be re-affirmed in
October, 1993 in order to remain an active standard.)
So, why all the fuss about calling Fortran 8x FORTRAN Extended?
Seems to me that if FORTRAN 77 is really contained in FORTRAN
Extended, then this issue should soon become a non-issue. FORTRAN
Extended compilers will appear on the market within the next 4 years.
Users will re-compile their code on them and execute it with no problems.
Most users will be so happy with the change to FORTRAN Extended that
there will be no interest in re-affirming FORTRAN 77.
Then FORTRAN 77 will remain as only a fond memory...brainerd@unmvax.unm.edu (Walt Brainerd) (09/02/89)
In article <1651@convex.UUCP>, psmith@mozart.uucp (Presley Smith) writes: > - Many things have been voted on in the X3J3 committee with "text to be > supplied later..." or "person x knows what to change". I know of NO > other committee that operates that way. Every other committee that > I've been a member of has operated ONLY on proposed changes with line > numbers and text for a particular document. We used to do this many years ago, but in recent years it has been policy to have a written proposal for every comma changed. Maybe this policy has been violated, but I can't remember it. > One example of the problem of getting the Fortran 8x document > correct has been on this net. The issue of removal of "with the > same interpretation" in TWO places in the document WAS DEFINED > in a DOCUMENT voted on by the committee which contained BOTH > PAGE and LINE NUMBERS to be changed for BOTH places in the > 8x draft, COMPLETE with the TEXT to be SUBSTITUTED in each case. > Due to human error or whatever, it was only changed in one > place. This seems like a trivial error. How many more serious > errors are there? > So even the suggested and followed method does not produce perfect results, as I suspect is the case in all other human endeavors. What is the point that you are trying to make? > > 2, The second point is that there's too much protest over keeping > keeping FORTRAN 77 as an active standard. If this is what you really want, what is wrong with having F77 be a subset of F8x, both recognized standards and described in X3.9-19??. -- Walt Brainerd Unicomp, Inc. brainerd@unmvax.cs.unm.edu 2002 Quail Run Dr. NE Albuquerque, NM 87122 505/275-0800
khb@road.Sun.COM (Keith Bierman - Advanced Languages - Floating Point Group ) (09/02/89)
In article <1651@convex.UUCP> psmith@convex.com (Presley Smith) writes: > > So, why all the fuss about calling Fortran 8x FORTRAN Extended? > Seems to me that if FORTRAN 77 is really contained in FORTRAN > Extended, then this issue should soon become a non-issue. FORTRAN > Extended compilers will appear on the market within the next 4 years. > Users will re-compile their code on them and execute it with no problems. > Most users will be so happy with the change to FORTRAN Extended that > there will be no interest in re-affirming FORTRAN 77. > Why is the current situation different than during the transition from X3.9-196x to X3.9-197x ? Now we fear codes will break, then the transition forced them to (hollerith data). Despite the '66 standard not having existed as such for many years, folks sucessfully run them ... both on f77 platforms and on machines with the old standard faithfully (more or less) maintained. You have advanced cogent arguments against large projects trying to convert overnight with the first f8x compilers to appear on the market. You have explained carefully how errors may have crept into the document. In a future posting would you please explain in what way the industry has changed .. so that it is now necessary to support an obsolete document, as opposed to the last "crank turn" of the language standard. Keith H. Bierman |*My thoughts are my own. !! kbierman@sun.com It's Not My Fault | MTS --Only my work belongs to Sun* I Voted for Bill & | Advanced Languages/Floating Point Group Opus | "When the going gets Weird .. the Weird turn PRO"
corbett@ernie.Berkeley.EDU (Robert Corbett) (09/02/89)
In article <318@unmvax.unm.edu> brainerd@unmvax.unm.edu (Walt Brainerd) writes: > >If this is what you really want, what is wrong with having F77 be a subset >of F8x, both recognized standards and described in X3.9-19??. > While I favor the idea of two separate standards, I would be willing to support the idea of a single standard PROVIDED 1) the full text of ANSI X3.9-1978 is kept unchanged, 2) the Fortran 8x extensions are added in the form of a supplement, and 3) an implementation is allowed to claim conformance to the FORTRAN 77 subset of the standard even if it contains extensions in conflict with the Fortran 8x extensions. Faithfully yours, Bob Corbett uunet!elxsi!corbett ucbvax!sun!elxsi.com!corbett corbett@berkeley.edu