How hard is it to revise an ANSI specification?

Also, if it was easy enough, what would some other drawbacks be to
revising the spec for CL?

The discussions around [url down]
and others bring into focus this question.

Timbo

På Tue, 23 Sep 2008 11:22:55 +0200, skrev macoovacany
<macoovacany>:

> How hard is it to revise an ANSI specification?
>

Very hard and very expensive. There is simply no incentive to do so among
the Lisp vendors.
But all is not lost. There is a project to provide standard extensions.
http://cdr.eurolisp.org/

--------------
John Thingstad

> There is simply no incentive to do so among  
> the Lisp vendors.
> But all is not lost. There is a project to provide standard extensions.[..]
>

So what was difference between circa CTL2 / ANSI specification vendors
and now? What was the original incentive?

Thanks for the link.

Timbo

macoovacany wrote:
>> There is simply no incentive to do so among
>> the Lisp vendors.
>> But all is not lost. There is a project to provide standard extensions.[..]
>>
>
> So what was difference between circa CTL2 / ANSI specification vendors
> and now? What was the original incentive?

http://www.dreamsongs.com/Files/Hopl2.pdf

Pascal:
I already had a copy, but had only read up to page 7. (D'oh). Just
quickly skimming it now...

([Page 20]: "Elitist Lisp"... hmmm. I think it was probably a good
thing that CL wasn't called this.
Flexures? Sounds like fun, but for another topic.)

OK. From what I can gather, the main motivating factor for the initial
set of meetings in the Spring of 81 was:

"If there were no consolidation in the Lisp community at this point,
Lisp might have died. ARPA was not interested in funding a variety of
needlessly competing and gratuitously different Lisp projects. And
there was no commercial arena—yet." [Top of Page 20]

So the difference between then and now was that ARPA was potentially
going to keep funding lisp. (Apologies for the over-simplification. I
have only skimmed it. Let me know about any other relevant factors.)

Whether or not there is a commercial arena... I guess there is.
Airline reservation systems, etc. (I don't need a list, it's on the
commercial vendor's site).

Hmmmmm. My conclusion is that CL has survived and has found a niche in
big projects. It's the small "make a web site in 30 minutes" kind of
projects that lisp is missing out on.

Seems a shame that the white, yellow and red pages mentioned didn't
make it. Would've made a good flag.

Timbo

In article
<21695d57-93e6-418e-a6df-1f7aeb82da8a>,
macoovacany <macoovacany> wrote:

> > There is simply no incentive to do so among  
> > the Lisp vendors.
> > But all is not lost. There is a project to provide standard extensions.[..]
> >
>
> So what was difference between circa CTL2 / ANSI specification vendors
> and now? What was the original incentive?

Money. Lots of money.

Much of the Lisp development in the 80s was paid for by the US Government
(mostly through DARPA, the defense research agency) and from companies making money
during the AI bubble. The AI bubble was comparable to the Internet bubble (which
then was a few years later).

Lots of universities and research groups were using
Lisp. The whole AI market was all in all a billion dollar market.
Symbolics for example had revenues over its lifetime of more than
500 Million dollars (which was more money than it is now, I'd guess).
Then there were companies like Xerox, DEC (remember them?), SUN,
Apple, TI and others that were active in this market. DARPA supported
the creation of Common Lisp and the ANSI Common Lisp standard. You
can see in that lots of people and companies were involved.
You can also say that DARPA demanded such a standard, since they
did not want to have the software developed in several incompatible
dialects. Lisp was seen a strategic tool to enable the creation
of intelligent machines and software (for DARPA: to improve
the capabilities to fight and win wars). Symbolics for example
sold lots of machines to the SDI program of Ronald Reagan's administration.
It was believed that the defense system needed extremely complex
software (like in the 100 Million lines of code range), that you
need tools for that and that Lisp was such a tool. The
US also reacted to the fifth-generation program from the Japanese
and started a research center (MCC) which was a heavy Lisp user
(for example to create VLSI design software, develop object-oriented
databases (Orion) and large-scale knowledge-based systems (Cyc)).

This is no longer the case that there is that amount of business. During
the 90s the market was very low. It increases a bit since 2000.

It took a long time to create the ANSI CL standard. A few things have changed
since then. There are no longer these deep pocket customers. The Internet
is now everywhere. Open Source is everywhere. Unfortunately ANSI has
not changed into something that would make sense to use for standardization - last I looked.
There is some historic baggage. For example that there are HTML versions
derived (!) from the ANSI CL standard was made possible by people
like Kent Pitman (and the vendor who paid for it). But still we can't just
take the ANSI CL document and create a new version - ANSI has the rights -
as I understand.

Just publishing CDRs is also not a simple solution to the problem. There is
nothing that motivates vendors (free and commercial) to implement it
or users to use it. You can see for example on the Scheme side that
the SRFIs are very successful (there are lots of them and people
have put a lot of work into them) - but then after years of SRFIs
you get something like R6RS where lots of people in the community
are not happy with.

I can imagine that with growing interest in Common Lisp there will
be some interest in a new standard which is based on existing
experience - and there is lots of experience. For example
conditions, pathnames, streams, ... could all be defined with
CLOS and get an extensible interface. Currently these extensions
(for streams: Gray streams or simple streams) are already
popular. But don't hold your breath.
[..]

På Tue, 23 Sep 2008 12:52:20 +0200, skrev Rainer Joswig <joswig>:

> dialects. Lisp was seen a strategic tool to enable the creation
> of intelligent machines and software (for DARPA: to improve
> the capabilities to fight and win wars). Symbolics for example
> sold lots of machines to the SDI program of Ronald Reagan's
> administration.

Funny I thought the millitary went for the TI-explorer.

--------------
John Thingstad

In article
<30555fea-0540-4f36-adab-9601ec17808c>,
macoovacany <macoovacany> wrote:

[..]
> commercial vendor's site).
>
> Hmmmmm. My conclusion is that CL has survived and has found a niche in
> big projects. It's the small "make a web site in 30 minutes" kind of
> projects that lisp is missing out on.
>
> Seems a shame that the white, yellow and red pages mentioned didn't
> make it. Would've made a good flag.
>
> Timbo

DARPA was a 'single' customer with a multitude of projects.
They also had a strategic program for creating a
whole landscape (hardware, software, ...).

Nowadays there are mostly customers with only a few projects.
For them basic interoperability exists and much more
is not needed - at least it does not help much to write
the vertical solution.

On 23 Sep., 13:01, "John Thingstad" <jpth> wrote:
> På Tue, 23 Sep 2008 12:52:20 +0200, skrev Rainer Joswig <jos>:
>
> > dialects. Lisp was seen a strategic tool to enable the creation
> > of intelligent machines and software (for DARPA: to improve
> > the capabilities to fight and win wars). Symbolics for example
> > sold lots of machines to the SDI program of Ronald Reagan's  
> > administration.
>
> Funny I thought the millitary went for the TI-explorer.
>
> --------------
> John Thingstad

DARPA wanted to create an ecosystem with several vendors. Symbolics
was
the biggest and DARPA helped TI to be a competitor. TI was
also wanting to get into the market, so they made low prices
to enter the market. Which was kind of problematic, because other
customers
also demanded lower prices from TI and Symbolics was happy
to point it out. ;-)

Having multiple vendors for hardware, software and services
was very important for DARPA.

macoovacany wrote:
> How hard is it to revise an ANSI specification?
>
> Also, if it was easy enough, what would some other drawbacks be to
> revising the spec for CL?
>

In Lisp, creating a new function/library is tantamount to extend the
language itself.
Therefore, an ANSI specification of a programming language does not
fit Lisp, and probably was not the right thing to do in the first
place.
Nowadays, it would suffice to have a central authority/mailing-list/
task-force to agree on names of functions given their described
behavior (I know, I know: functions, macros, variables...)

Disclaimer: This message is not meant to be offensive or even
provokative.

> In Lisp, creating a new function/library is tantamount to extend the
> language itself.
> Therefore, an ANSI specification of a programming language does not
> fit Lisp, and probably was not the right thing to do in the first
> place.

Ahhh. Completely agree.

In the PDF provided by Pascal, the first CL meeting started with a
comment about MacLisp going in four different directions. It seems to
be the easiest thing in the world to do in lisp: go off in your own
direction. I'd even call it a fundamental problem, but someone else
has already reserved that phrase. (Yes, a greatest strength can also
be a greatest flaw. Whole stories are based around that.)
The result of roll-your-own attitude to libraries is a perception of
abandware libraries (true or not is irrelevant. It's the PERCEPTION).

> Nowadays, it would suffice to have a central authority/mailing-list/
> task-force to agree on names of functions given their described
> behavior (I know, I know: functions, macros, variables...)
>

A central authority in Lisp? Any self-nominated ones will be
inevitably not be recognized by someone. Back to start.

The best so far has been Alexandria to keep in touch with supported,
reasonably cross compatible libraries.

Tim

macoovacany <macoovacany> wrote on Tue, 23 Sep 2008:
> How hard is it to revise an ANSI specification? Also, if it was easy
> enough, what would some other drawbacks be to revising the spec for CL?
> The discussions around
> [..] and others
> bring into focus this question.

You know, what's odd about your claimed motivation, is that the referenced
articles complains about everything about Lisp _except_ the spec. The point
of the article is that the language is great, the spec is great ... but for
social and practical and libraries and popularity reasons, the author
believes Lisp is a poor choice for web programming.

If that article motivated you, why would it motivate you to revise the
CL ANSI spec? That was the only thing the author _didn't_ complain about!

-- Don
__________________________________________________ _____________________________
Don Geddis http://don.geddis.org/ don
When Bob asked me to watch his house while he was gone, that's what I did. He
didn't say anything about calling the fire department.
-- Deep Thoughts, by Jack Handey [1999]

macoovacany <macoovacany> wrote:
+---------------
| Hmmmmm. My conclusion is that CL has survived and has found a niche in
| big projects. It's the small "make a web site in 30 minutes" kind of
| projects that lisp is missing out on.
+---------------

Possible counter-examples?

http://rpw3.org/hacks/lisp/minimal.lhp
http://rpw3.org/hacks/lisp/appsrv-demo.lhp


-Rob

-----
Rob Warnock <rpw3>
627 26th Avenue <URL:http://rpw3.org/>
San Mateo, CA 94403 (650)572-2607

rpw3 (Rob Warnock) writes:

> macoovacany <macoovacany> wrote:
> +---------------
> | Hmmmmm. My conclusion is that CL has survived and has found a niche in
> | big projects. It's the small "make a web site in 30 minutes" kind of
> | projects that lisp is missing out on.
> +---------------
>
> Possible counter-examples?
>
> [..]
> [..]

It's more a matter of this:

% sudo apt-get install libapache2-mod-php5
% mkdir public_html
% nano public_html/index.php

Getting from a fresh operating system installation to the point where
you can start editing index.lhp is a bit more involved. That's where
you burn up your 30 minutes.

On Sep 24, 1:53 am, Don Geddis <d> wrote:
> macoovacany <macoovac> wrote on Tue, 23 Sep 2008:
>
> > How hard is it to revise an ANSI specification?  Also, if it was easy
> > enough, what would some other drawbacks be to revising the spec for CL?
> > The discussions around
> >[..] others
> > bring into focus this question.
>
> You know, what's odd about your claimed motivation, is that the referenced
> articles complains about everything about Lisp _except_ the spec.  The point
> of the article is that the language is great, the spec is great ... but for
> social and practical and libraries and popularity reasons, the author
> believes Lisp is a poor choice for web programming.
>

Yes. And why would lisp have social and practical and libraries and
popularity reasons to make it poor choice for web programming?

I think the answer is too many choices.

Why are there so many choices?
Because there is not 'only one' choice. There are many choices.

Why are there many choices?
Because CL is a Spec, not a single implementation.

Can we get rid of the Spec and have a single provider? (For
completeness only)
No. There are a number of companies that like things the way they
are.

Can we make a single implementation (practically, if not officially)?
Adoption of an accepted libraries for computing stuff (sockets, web
authorings, etc) that was not part of the original specification.
(Yellow pages? Alexandria)

Can we make this a de facto standard?
NFI.

Can we revise the spec?
The original spec cost ~400K, when turnovers for companies were ~500M.
Those figures aren't happening now. Neither is a Spec revision.

Should we revise the spec.
Topic for another thread.


**************************

Where abouts do we disagree on this train of thought?

Tim