Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!swrinde!sdd.hp.com!spool.mu.edu!uunet!tut.cis.ohio-state.edu!ucbvax!Eng.Sun.COM!Mitch.Bradley
From: Mitch.Bradley@Eng.Sun.COM
Newsgroups: comp.lang.forth
Subject: Re: What should the Standard include?
Message-ID: <9105081638.AA10000@mitch.Eng.Sun.COM>
Date: 8 May 91 16:38:51 GMT
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The Internet
Lines: 86


> All I can say is, the Forth community must be a lot smaller than I thought
> it was.  (Someone once told me that there were about 150 Forth vendors.)

In the early phases of the standards effort, the TC sent out hundreds
of copies of a questionnaire.  They tried to send it to every vendor they
could locate.  The questionnaire was fairly detailed, asking for information
about product's compliance and areas of noncompliance with existing standards,
areas where the standard should be improved, and other things.  It also
asked for copies of product manuals so the TC could learn details of these
implementations.

The response was underwhelming.  The TC got very few responses back.

> In fact, the big weakness of my proposed criteria is
> that they allow all six model systems to come from a single vendor.

I presume that citing 6 systems from the same vendor would not satisfy
the spirit of your requirement.  If so, then this criterion would be
useless; I could justify anything I wanted based on different versions
of Forthmacs.

> Specific examples?
> ONLY/ALSO.

Maybe, but there are serious problems with the semantics of VOCABULARY
and differences in the implementation of SEAL .  Also, in order to
make the ALSO/ONLY wordset complete, you need PREVIOUS, which probably
wouldn't "make the cut", and you also need some way to save and restore
the search order, which wouldn't make the cut.

>  TO.

Maybe.  It depends on how strictly you construe the "no semantic differences"
requirement.  Also, TO is only useful in the context of VALUE , which might
not satisfy the requirement, and with locals, which definitely would not
satisfy the semantic requirement.

> Conditional compilation.

Semantic differences exist.

> FOR...NEXT.

I doubt it.

> Probably file systems

I'm quite certain that it would not have been possible to achieve a
coherent, useful file system wordset based upon your proposed criteria.

> Probably much of floating point (ditto).

Yes, I expect that much of the floating point wordset would have passed.
But what about the rest?  A wordset with gaping holes is crippled.

> Almost certainly you could find two
> or three strings packages to meet these criteria.

Perhaps.  The only example of a strings package that would come close
is FSTRINGS, and I don't know of any Forth systems that come with it.

> And given the number of vendors on the TC, it should take no more than 12
> months and a fraction to meet this requirement for the most fevered products
> of the TC's imagination.  Why is no one willing to do this?

There are 4 significant vendors represented on the TC:
        FORTH, Inc (polyFORTH), Creative Solutions (MacForth),
        Bradley Forthware (Forthmacs), Vesta (some ROM Forths)

LMI (UR/Forth) used to be represented.  In addition, we know pretty much
about F83, F-PC, and MVP Forth from manuals and personal experience.

There is very little agreement between the products of these vendors in
terms of detailed semantics of extension packages.

> The burden of proof is on the positive, in philosophy and in
> jurisprudence....but evidently not in standards-making.

Philosophers can invent their own data.  In jurisprudence, what it really
comes down to is what the jury votes.

The ANS Forth committee relies on judgment, debate, and voting to make
its decisions.

Mitch.Bradley@Eng.Sun.COM