Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!csd4.milw.wisc.edu!lll-winken!uunet!seismo!sundc!pitstop!male!sun!fatcity!khb
From: khb@fatcity.Sun.COM (Keith Bierman Sun Tactical Engineering)
Newsgroups: comp.lang.fortran
Subject: Re: Responses to M. Shapiro & K. Bierman, re: X3J3 and WG5 (long)
Keywords: WG5, X3J3, F8X draft
Message-ID: <98161@sun.Eng.Sun.COM>
Date: 9 Apr 89 23:09:12 GMT
References: <24091@beta.lanl.gov>
Sender: news@sun.Eng.Sun.COM
Reply-To: khb@sun.UUCP (Keith Bierman Sun Tactical Engineering)
Organization: Sun Microsystems, Mountain View
Lines: 137

In article <24091@beta.lanl.gov> dd@beta.lanl.gov (Dan Davison) writes:
>
>Keith Bierman says:
>
>> I suggest you study the last 10 years of meeting notes. It is not a CS
>language product.
>
>How would an outsider get copies of these?

become an X3J3  observer. For $150ish/year you get everything
published by the committee. There is probably no formal mechanism for
procuring the entire collected back copies, but you can nag the old
time members.... :>

>
>>Inasmuch as WG5 "held out" for a very short list of features, in a
>>very large document, it seems quite unreasonable to bitch about WG5
>>running the show.
>
>Waitaminnit.  My comments were based on the only information I had, the
>Digital News article.  Your comment does not jibe with the article.

No. I understand what I'm talking about. 

>
>I now have a copy of the draft, and with the '77 standard in hand,
>and the old 8x standard, I plan a cozy weekend of reading it over.
>Hopefully all my perceptions are wrong.
>
>BTW, there are others out there whose *perceptions* match mine.  
>Perhaps the confusion over the last document had something to do 
>with its presentation.
>

Much did. Walt Brainard, does seminars. Jeanne Adams (chair) and J.
Wagoner (vice chair) have given tons of them. Learning the language
from the standards document is not a pleasant process. Alas, the
standards document MUST address the implementors, and not the users.
In europe there was much greater flexibility in presentation. ANSI
rules are more than a little constrictive.

>>Also note that you have had over a decade to voice your views.
>
>Not me, I've only had the money to pay for the draft in the last two
>years. That's not exactly encouraging comment.
>

The speakers listed above (and many others from the committee) work
for users (ncar, ammocco, university, for starters) and have been
speaking with, and for their user communities. If your organization
isn't represented, you can't blame the committee! All meetings are
open to the public. 

>>Modulo the misinformed comments
>
>Perhaps if ANSI cared enough to make proposed standards available at
>a sensible cost, you'd get the informed comment you prefer.  It's not
>X3J3's responsibility, but complaining seems a little out of place.

ANSI rules specifiy what the committee is allowed to do. ANSI uses the sale of
standards documents to finance itself. The european equivalents have
government support, so they feel differently about pricing stuff.

We understand that this is very counter-productive, and we are working
to find a way around the ANSI rules. But it is difficult.

>.....
>>>Not to me.  I will read the proposed standard with a fairly open mind,
>>>but it appears that we will get the European standard instead of a
>>>US standard.
>
>>Since we (americans) wrote the document, and continue to do all the>
>>technical work, this is simply silly.
>
>Ahh, semantics.

No. Those who write the document have the most control. Working from
memory (which is quite fallible) WG5 held out for :

1)	pointers, with a certain syntax

	(* which many americans asked for in public comment *)
	Pointers destroy 90% of optimization....we don't want f8x to
	be as non-optimizable as c. The user community has spoken,
	they WANT to give up optimization. The best we (vendors) can
	do is teach them NOT TO CODE WITH THEM.

	It is far from clear that the "european" syntax is better/worse
	than the "american" syntax. The "european" proposal was, in fact,
	written by american committee members.

2)	some misc "spelling" changes

3)	4, 5  escapes me this moment.

From where I sit, the pointer debate was the biggest difference
between the US and europeans. The committee debated the issue at
length over a period of years (long before I got there). And the
forces of optimization had won out. The american public comment, and
the euopean comment went strongly in favor of pointers. The syntax
selected has some chance of allowing optimization (unlike c).
>
....
>>entities are multiplied. 
>
>Amen. I much prefer having the compiler figure out the most efficient
>ordering of a matrix multiplication.  There are a phenominal number
>of ways to do it badly on a Cray.

Welcome f8x.

>
>>The proposed standard will allow me to write code which will 
>>be easier to read, port, maintain and tune. This is what I DEMAND 
>>from a language for scientific applications. 
>
>This is what I *wish*, or just give up  and use APL.  The problems
>with APL write-once-APL are legendary.
>

f8x is much better than APL. Before joining the vendor side, I was
involved in many Kalman filtering projects. With F8x, I can easily
create a library which knows about triangular matrices, and hides the
implementation from users...they will mainly write

	U = PHI * U

never knowing (or caring) that phi is square (or triangular), and that
u is always triangular (with D stored in the diagonal). 

Cheers.




Keith H. Bierman
It's Not My Fault ---- I Voted for Bill & Opus