Re: [Axiom-developer] Aldor and Lisp

[Camm Maguire]

Greetings, and thank you for this initiative!

Martin Rubey <address@hidden> writes:

> I intend to send the following text to the lisp newsgroup today
> afternoon. Comments welcome.
> 
> Dear all,
> 
> I'm one of the developers of the Axiom project, a pretty sophisticated --
> meanwhile free -- Computer Algebra System. One of the distinguished features 
> of
> this system is it's extension language, which is -- in my opinion -- extremely
> well designed.
> 
> However, there is a slight problem. Some years ago, this extension language --
> than a commercial product -- was redesigned and in the process rewritten in
> C. The new language was called Aldor.
> 
> Time passed by and Aldor became something gratis, but unfortunately, not quite
> free. In particular, this means that the source cannot easily be obtained and
> there is very little development on the source itself.
> 
> One important point is that at that time, Axiom (the computer algebra system)
> was not free at all and available only to few people. Meanwhile, this has
> changed, and it seems that there is quite a lot of interest in it. Bugs get
> fixed and great contributions make the system become partly superiour to the
> big players. In fact, you can try it out online:
> 
> http://page.axiom-developer.org/zope/mathaction/FrontPage
> 
> Although Aldor can be used as the extension language for Axiom, some of its
> most remarkable features cannot, since they lack support from Axiom.
> 

If someone can concisely specify exactly what is missing in SPAD vis a
vis Aldor, it would be most helpful.  What are these most remarkable
features?  Examples?  Alas, I lack time to plough through the 300 page
aldor manual.

Just my 2 cents here -- it really doesn't matter what tool we use, as
long as it is open source and we can get mind-share circled around its
development.  I know this first hand in working with GCL from scratch
knowing no lisp at all.  Once one decides to invest in the learning
curve, the very last thing we should do is flip flop -- expertise is
the value we supply and it takes time and is hard to come by.  Given
where the cards appear to lay right now, extending/fixing SPAD to
match Aldor appears the obvious choice.  I am assuming a lot here, but
it is primarily that the factors behind the Aldor rewrite decision
went something like 1) this code has cruft, time for a rewrite 2)
we're in the AI winter, nobody uses lisp anymore, and its slow, 2.5)
its more fun and easier to start over than try to understand and build
upon the work of others, even if it does effectively destroy the value
of everyone's time spent on the older stuff, 3) lets build something
totally new in C.  These motivations are not commensurate with those
we face in the open source world.  I'm also assuming that Aldor is
basically like spad but nicer, faster, and a little more powerful.  If
I'm missing something as is quite likely, the conclusion obviously
does not hold.

> -------------------------------------------------------------------------------
> 
> So why am I posting to the Lisp community? 
> 
> The reason is, that Axiom is written in Lisp -- it currently runs on
> GCL. (Porting to "other common lisps" would also be greatly appreciated...)
> 

Here too, it is a complete waste of time, IMHO, trying to make sure
axiom or any system runs on 11 different compilers.  Choose one,
establish the expertise needed to ensure that it will remain
servicable for the indefinite future, and then concentrate on adding
functionality to AXIOM.  

I've been having a lot of correspondence lately with theorem-prover
authors, and I've come to appreciate Tim's original idea of using one
to verify AXIOM.  Having something like this would clearly put AXIOM
at the head of the pack IMHO.

> I believe that it would be wonderful to have an Aldor compiler and interpreter
> written in Lisp.
> 
> Of course, the current compiler and interpreter for Axiom's extension language
> ("spad"), is written in a dialect of lisp (called "boot"), which itself is
> written in lisp.
> 
> The greatest thing would be to enhance the present spad compiler and
> interpreter to be compatible with Aldor. The semantics of Aldor are available

Again, specifics here would be most helpful.  I don't think anyone has
time for the document below.

> online, see Chapter 22 of 
> 
> http://www.aldor.org/docs/aldorug.pdf.gz
> 
> and do not differ greatly from the semantics of spad.
> 
> A milestone would be to enable the Axiom interpreter to use the features
> provided by Aldor -- in fact, this boils down to *one* feature that is 
> missing.
> ("dependent signatures")

Intriguing -- more here please?  Sounds like some kind of ML like
feature, which I have been thinking about in other contexts recently. 

Take care,

> 
> -------------------------------------------------------------------------------
> 
> If you are interested to work on this subject, please join us! There is a lot
> more information available, and quite a bit of work has been done already. The
> appropriate mailing list is address@hidden
> 
> Martin Rubey
> 
> 
> 
> _______________________________________________
> Axiom-developer mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/axiom-developer
> 
> 

-- 
Camm Maguire                                            address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah