(j3.2006) For consideration at 200

[Van Snyder]

On Wed, 2012-11-28 at 07:52 -0600, Bill Long wrote:
> On 11/27/12 9:00 PM, Van Snyder wrote:
> > In addition to the proposal for a TS on units, my sponsors have asked me
> > to put forward 21 proposals, of which 15 are described by papers from
> 
> I get a lot of requests from "my sponsors" for new features, often aimed 
> at a performance tweak for a  code that is either archaic or was not 
> very well written in the first place.

That programs might not have been "well written in the first place" is a
pretty lame excuse for not making it easier to improve and maintain
them.  Maybe one of the reasons they were "not well written in the first
place" was that the only tool available was FORTRAN II or Fortran 66.
One does not glibly or eagerly start over on a six million line
mission-critical code, that has been under continuous development for 53
years and now works remarkably well, just because the tools to write it
well in the first place did not exist "in the first place."

If JPL were the only institution having legacy Fortran code that is
important to maintain and improve, there would be little reason to worry
about compatibility of revisions of the standard.

My sponsors are more interested in labor cost and reliability than minor
performance tweaks.  We considered coarrays to be a win on all three
counts: lower labor cost, increased reliability, and (perhaps
substantially) increased performance.  We get minor performance tweaks
just by waiting for a few months to buy a new processor.

> I see part of my J3 role as 
> filtering ....  Perhaps it would be 
> beneficial for all concerned if you were more aggressive in explaining 
> to your sponsors why some of these proposals are unlikely to pass.

Compare the size of Tutorials/JPL-wishlist-2012.pdf on the server, or
maybe only its three-page table of contents, to the number and scope of
projects I am presently proposing.  There are 117 items in the TOC, and
some of them might be considered to represent several projects.  Are you
really serious that I'm not filtering proposals?

I'm having trouble explaining to my sponsors why it is valuable to them
for me to participate in Fortran standardization since nothing at all
happened concerning gathering requirements for the next revision during
the last four years, nothing of significance will be done in the next
revision, and none of the few proposals I have not talked them out of
appear in the minutes.  They're wondering why they sent me to Las Vegas
14 times, and halfway around the world twice, to accomplish... nothing.
Perhaps that is part of the reason that J3 has only three user members:
maybe other potential members have the same problem.

> I'm not completely opposed to new features for f201x.  But the feature 
> needs to be small, and trigger a reaction of "hey, that is a really 
> good, useful idea" or "we probably should have done that in the first 
> place".

So let's have an e-mail discussion and maybe an informal letter ballot
on the size, merit, and appropriateness of my proposals, so as not to
"waste" J3 time, preferably after the papers are available, instead of
ruling apriori that they're out of order before they appear.  Does
anybody want to discuss the size and content of the proposals before
February, or will the first order of business at 200 be to rule them out
of order without looking at them?

One of my proposals is, in fact, of the form "we probably should have
done that in the first place".

All I'm asking for is a fair hearing, and a result that is more decisive
and convincing than a subgroup vote of 1-1-1.  I didn't complain about
the process or outcome in 2004, when my proposals were fairly and
rationally considered, and most of them rejected or tabled in favor of
others that had higher priority.

> If I though it was open season on proposals the size of Units,

I am not proposing Units as a project for the next revision.  I am
proposing it as a TS, with a schedule entirely independent from any
other J3 or WG5 project.  I have been working on units for 34 years.  I
have been working on the present proposal document for four years.  It
is in far better condition than the Coarray TS, or the vulnerabilities
TR, or the first several revisions of the interop TR.  I will not be
asking J3 to write it for me.

Comments on the content of the units proposal (or indeed any of my
proposals) are welcome.  The more that can be done offline, the less J3
time they take at meetings.

> We already have two 
> big features - the interop TS and the coarray TS. And there is a lot of 
> work left to be done on the latter.

Hopefully, J3 and WG5 "online" time will mostly be spent on gathering
and deciding requirements.  Writing a TS in plenary is a very
inefficient use of our time.