---------- Forwarded message ----------
From: Bill Page <bill.page@newsynthesis.org>
Date: Fri, Mar 21, 2008 at 5:07 PM
Subject: Re: What Axiom can do for Aldor
To: William Sit <wyscc@sci.ccny.cuny.edu>

William,

With your permission, I would also like to post this entire email to the blog on Algebraist. Would that be ok?

Of course since this is also directly related to Aldor it would certainly be ok also for you to reply via the aldor email list (the old fashioned way :-).

On 3/21/08, you wrote:
>
> I don't know what I am getting into! My "network" so far
> has nobody in it. I have "facebook"-phobia.
>

No problem. So far you have become a member of network (called Algebraist) because you replied to my invitation. Thanks for re-registering with your personal email address. I have deleted your previous registration that included the axiom email list address.

Since I created Algebraist, when you responded to my invitation you were also automatically added to my own personal "network" of friends. Now if you want to, you can click on the profiles of any of the members of Algebraist and mark them as *your* friends, i.e. part of your "network". This is entirely optional, but I guess it is part of the "fun" of this type of web system.

I have to admit that I am also rather new to this "social network" thing on the web. I have never used facebook and also have a kind of phobia about even trying it. I guess this happens to all of us as we get older ... :-) But the 'crowdvine.com' system struck me as a little different (easier and less scary) since I first encountered it as part of the organization for a conference. So when I was setting up the parallel development environment for Aldor at http://algebraist.origo.ethz.ch/ I decided that crowdvine might also serve quite well as a "front-end" to that site.

> At this time, I'm "retired" but I have actually been quite
> busy (more than I want to be) in organizing a seminar, two
> conferences, editing two proceedings and editing a book by
> my wife (the killer for this one is the indexing and page
> breaks). So I have learned to gradually avoid future
> commitments. I can only offer little to this new project
> of yours. Plus, I am totally unfamiliar with Aldor and
> very confused with the flavors of Axiom (but that may be
> because I have not joined the mailing list for FriCAS or
> OpenAxiom).
>

I agree the situation remains confusing. I had rather hoped that two forks (OpenAxiom and FriCAS) would merge into one, although I had little hope that there would be any reconcilation with Tim's original Axiom project. But I think that very gradually the three Axiom-related projects have started to stake-out their own "territories":

Tim's Axiom - still primarily documentation plus MathML browser-based replacement for hyperdoc

OpenAxiom - enhancements focusing on improving and extending SPAD, although not ncessarily in the same direction as the Aldor project. Gaby's interests seem to be strongly motivated by his background and teaching responsibilities in compiler design even though the project goals seem more ambitious and largely overlap FriCAS.

FriCAS - focused primarily on fixing bugs in the original Axiom system at both the system (hyperdoc is much more reliable in FriCAS) and the algebra library level (with Martin Rubey is working on fairly deep changes to the library). Also support for more alternate Lisp and operating system environments. FriCAS is the first full version of Axiom that will build on Windows (cygwin).

There has been some effort to migrate changes from one version to the another (such as those algebra fixes not requiring major changes in the database and the new broswer interface) but of course there are not enough resources to spare to make sure that this happens for each major improvement.

>
> As to your two points of view: What Axiom can do for Aldor
> and what Aldor can do for Axiom, I actually don't see much
> of a difference in the sense that I think the best is to
> merge the two into one. As long as technically this is not
> feasible (for lots of reasons perhaps), I would prefer to
> stay with Axiom, for the reason that I am more familiar
> with it and hence it would be the more useful to me.
>

Actually, I think it is both technically and legally feasible *provided* we are willing to adopt the (semi-)open source license for the entire package. It would then be possible to distribute a version of Axiom along with Aldor in installable binary form that has Aldor already built-in to the package, rather than struggling with a separate build step for the Axiom/Aldor interface (Peter Broadbery's work). It is complicated somewhat now by that fact that the directory structures of the Axiom forks have diverged somewhat - so one "size" of build system no longer fits all... :-( But this would be solved (at least for Aldor users) by including some version of Axiom with the Aldor distribution and providing an integrated makefile that builds the system Axiom+Aldor as a whole.

> You may ask then why did I join. Well, I trust whatever
> you are doing and so I'll try to stay informed.
>

:-) I appreciate you confidence. I am very glad you joined and I look forward to your contributions to the discussion. And of course I do hope you find it informative.

> Regarding Aldor, clearly the founders of Axiom saw the
> futility of maintaining Axiom (or the compiler) and opted
> for a new language in Aldor. The problem they left us is
> that much of the Axiom library had to be rebuilt into
> Aldor, and updated. If I am not mistaken, this has not
> been completed (may be only 10%? That's a wild guess). On
> top of this, the Daly requirement to document all code
> meticulously (instead of leaving the "obvious"
> undocumented) increases the dullness (non-glamour) of the
> job. But you know all this.
>

Indeed. I agree with your summary.

> You seem to be more in favor of Aldor ("What Axiom can do
> for Aldor") than Daly ("What Aldor can do for Axiom").
> Along this line, one idea would be to write a
> (bi-)conversion program: this actually must have existed,
> come to think of it, and so what really is the problem
> with porting Axiom libraries to Aldor?

There is a program in Axiom that is intended to facilitate the conversion of SPAD to Aldor code but the conversion is incomplete and does not attempt to resolve any deep incompatibilities.

> Is there some inherent backward incompatibility?

Yes.

> Can we identify these?

I believe so.

> How can we make Aldor totally backward compatible without
> lessening the Aldor language?

Perhaps some of the ideas that Gaby is working on in OpenAxiom will help.

> Can we port the interpreter to Aldor?

Yes, I think so. But it would be a big job. Maybe a designing a new user interface written in Aldor would be of greater interest.

> Hyperdoc? Graphics? User-interface (display technology)?
>

Yes. As I mentioned above there has been some work (mainly by Arthur Ralfs and Tim Daly) on replacing hyperdoc with an interface to a MathML-capable browser (FireFox). There has also been some work by Waldek on making the graphics code more portable, e.g. to cygwin on Windows.

All of this goes quite far beyond what is available now in Aldor's native interpreter mode (e.g. Aldor -gloop). So although there has been a trend in the last 5 (or now almost 10) years to development stand alone libraries in Aldor, I think that Axiom as a "mature" computer algebra system environment is still a better place to test and develop new algebra libraries in written in Aldor. The difficulty has been that Axiom's own library sometimes gets in the way. Until the improvements to the build system developed by Waldek which now allow the entire Axiom library to be bootstrapped from an empty database, it was very difficult for Aldor library users to avoid this problem.

> I'll be meeting with someone familiar with Aldor tomorrow
> and I'll see what I can learn from him.

Great.

> However, as I will be on a trip for the next 10 days after
> Sunday, I won't be able to write you till I return.
>

No problem. I hope you have a good trip.

Regards,
Bill Page.