Axiom for Aldor or Aldor for Axiom?

[Bill Page-7]

Dear Aldorados,

(:- "Aldorado" = heterographic homophone of Eldorado i.e. "the Golden One" :-)

During the recent Aldor workshop the subject of the relationship
between Aldor and Axiom was discussed. Some of the participants
expressed the point of view that Axiom was important to Aldor
developers as an environment for testing complex math libraries such
as BasicMath. In general the interactive 'aldor -gloop' mode was
considered too primitive for this purpose, but Axiom fits this role
almost perfectly because it includes other library functions that can
be very useful for the development and testing new libraries. Whether
the new Aldor library is eventually intended to be used stand alone or
as part of Axiom is usually not so important. But what is important is
the development environment provided by Axiom.

This idea came as a bit of a surprise to me since normally I had been
thinking of Aldor as a missing and essential part of Axiom - the next
generation library compiler. But I realize now that this might work
both ways. Of course there are alternatives - such as implementing a
more powerful and convenient interpreter directly in Aldor - but this
is probably best viewed as a longer term goal which is more easily
obtained right now directly from Axiom.

So the issue that arises: Now that Aldor is available in an open
source form, how best to package Aldor and Axiom to facilitate this
use? Currently Aldor's APL2 source license prohibits commercial use of
the Aldor source code. This apparently makes it incompatible with GPL.
However it might be important to observe that software that is
licensed under modified BSD is both GPL and APL2-compatible - meaning
that software licensed under modified BSD could be combined with
Aldor's APL2 and released as one product under APL2. Axiom is an
example of such a package.

Although the comments in this email are not intended to be primarily
about licensing issues, I can well imagine that because it might
divert some resources away from Aldor, aldor.org might be reluctant to
package Axiom in this way specifically for use with Aldor. Maybe I am
wrong and I would like to hear more opinions about this from Aldor
developers. Similarly the current Axiom projects (including FriCAS and
OpenAxiom) might be equally reluctant to adopt the non-commercial use
clause of APL2 since that would make them incompatible with GPL.

So what (if anything) should we (both Axiom and Aldor developers) do
about this? Certainly with three Axiom forks already, it might stretch
credibility to far to fork Axiom again with an APL2 license just to
pursue the original goal of re-implementing the Axiom library in
Aldor. Is there an existing Axiom project/fork that is willing to
adopt Aldor as the new library compiler?

I think that from a technical point of view right now the FriCAS fork
of Axiom might be the most suitable system in which to re-implement
Axiom libary code in Aldor since FriCAS now builds both the algebra
code and the associated databases from directly from source using a
new bootstrap procedure. This makes it possible to make deep changes
in the Axiom algebra code without encountering Spad errors due to
inconsistent database files. But Gaby (the primary developer of
OpenAxiom) has also stated that it is the intention of the OpenAxiom
project to pursue similar goals as the original Axiom developers. Does
this include the goal of re-implementing (all or some part of) the
Axiom library in Aldor?

Regards,
Bill Page.

_______________________________________________
Aldor-l mailing list
Aldor-l@...
http://aldor.org/mailman/listinfo/aldor-l_aldor.org

[Gabriel Dos Reis-2]

"Bill Page" <bill.page@...> writes:

| Dear Aldorados,
|
| (:- "Aldorado" = heterographic homophone of Eldorado i.e. "the Golden One" :-)

Bill --

  Thanks for the report!

[...]

| So the issue that arises: Now that Aldor is available in an open
| source form, how best to package Aldor and Axiom to facilitate this
| use? Currently Aldor's APL2 source license prohibits commercial use of
| the Aldor source code. This apparently makes it incompatible with GPL.
| However it might be important to observe that software that is
| licensed under modified BSD is both GPL and APL2-compatible - meaning
| that software licensed under modified BSD could be combined with
| Aldor's APL2 and released as one product under APL2. Axiom is an
| example of such a package.

There is a little snag.  Axiom currently uses GCL, which by its
licensing -- if I understand correctly -- makes the end product GPL,
for the components BFD and UNEXEC.  This, I believe, is a consequence
of the Lisp technology itself -- you don't use it; it invades you.

A way out might be for Axiom to use SBCL which is in public domain.

[...]

| I think that from a technical point of view right now the FriCAS fork
| of Axiom might be the most suitable system in which to re-implement
| Axiom libary code in Aldor since FriCAS now builds both the algebra
| code and the associated databases from directly from source using a
| new bootstrap procedure. This makes it possible to make deep changes
| in the Axiom algebra code without encountering Spad errors due to
| inconsistent database files. But Gaby (the primary developer of
| OpenAxiom) has also stated that it is the intention of the OpenAxiom
| project to pursue similar goals as the original Axiom developers. Does
| this include the goal of re-implementing (all or some part of) the
| Axiom library in Aldor?

It is OpenAxiom's intent to inter-operate with libraries written
Aldor.  However, at this moment, I do not anticipate a complete
rewrite of OpenAxiom's algebra in Aldor.  One thing at the time :-)

I would like, first, to improve inter-operability with Aldor libraries
from within Axiom (and vice versa)  through ISA like FOAM.

-- Gaby

_______________________________________________
Aldor-l mailing list
Aldor-l@...
http://aldor.org/mailman/listinfo/aldor-l_aldor.org

[Bill Page-7]

On 03 Sep 2007 21:43:53 -0500, Gabriel Dos Reis wrote:
> ...
> It is OpenAxiom's intent to inter-operate with libraries written
> Aldor.  However, at this moment, I do not anticipate a complete
> rewrite of OpenAxiom's algebra in Aldor.  One thing at the time :-)
>

The last commercial version of Axiom was shipped with some Axiom
packages and domains already converted to Aldor. I think it would be
good to continue this conversion process in an incremental manner.

> I would like, first, to improve inter-operability with Aldor libraries
> from within Axiom (and vice versa)  through ISA like FOAM.
>

Sorry, what is "ISA"?

Are you aware that currently Aldor works within Axiom by compiling to
Lisp? How do you see FOAM being involved? Could you expand a little on
this comment?

Regards,
Bill Page.

_______________________________________________
Aldor-l mailing list
Aldor-l@...
http://aldor.org/mailman/listinfo/aldor-l_aldor.org

[Gabriel Dos Reis-2]

On Mon, 3 Sep 2007, Bill Page wrote:

| > I would like, first, to improve inter-operability with Aldor libraries
| > from within Axiom (and vice versa)  through ISA like FOAM.
| >
|
| Sorry, what is "ISA"?

"Instruction Set Architecture".  This is in reference to my goal to
have a well defined Axiom Virtual Machine that is not tied to Lisp.

| Are you aware that currently Aldor works within Axiom by compiling to
| Lisp?

Yes. But, eventually, I want to move away from Lisp.

| How do you see FOAM being involved?

I would like to see more documentation of FOAM and its semantics -- I
know there is a documentation; but I'm not sure how current it is and
whether it provides me with all I need.  I briefly discussed the
documentation issue with Stephen when I was in London last July.  


-- Gaby

_______________________________________________
Aldor-l mailing list
Aldor-l@...
http://aldor.org/mailman/listinfo/aldor-l_aldor.org

[Stephen Watt]

On Mon, Sep 03, 2007 at 11:22:31PM -0400, Bill Page wrote:
> ...
> Are you aware that currently Aldor works within Axiom by compiling to
> Lisp? How do you see FOAM being involved? Could you expand a little on
> this comment?

In case it is not clear:  FOAM is the compilation target for Aldor, and this
is used to generate low-level Lisp.  

FOAM is designed to have only operations that are efficient both in Lisp
and in a machine code-only environment.  (So, for example, there is no way
to take an address of something.)

-- Stephen

_______________________________________________
Aldor-l mailing list
Aldor-l@...
http://aldor.org/mailman/listinfo/aldor-l_aldor.org

[Bill Page-7]

On 9/3/07, Stephen Watt wrote:
> On Mon, Sep 03, 2007 at 11:22:31PM -0400, Bill Page wrote:
> > ...
> > Are you aware that currently Aldor works within Axiom by compiling to
> > Lisp? How do you see FOAM being involved? Could you expand a little on
> > this comment?
>
> In case it is not clear:  FOAM is the compilation target for Aldor, and this
> is used to generate low-level Lisp.
>

Yes, I think that is clear. I wanted to understand how FOAM could be
used in Axiom as implied by Gaby's email.

> FOAM is designed to have only operations that are efficient both in Lisp
> and in a machine code-only environment.  (So, for example, there is no way
> to take an address of something.)
>

As an abstraction layer this makes very good sense. I believe that
Aldor also provides an interpreter for this layer: "-gloop", right?
But it is not clear to me how Axiom could make use of this.

On 9/3/07, Gabriel Dos Reis wrote:

> On Mon, 3 Sep 2007, Bill Page wrote:
>
> | > I would like, first, to improve inter-operability with Aldor libraries
> | > from within Axiom (and vice versa)  through ISA like FOAM.
> | >
> |
> | Sorry, what is "ISA"?
>
> "Instruction Set Architecture".  This is in reference to my goal to
> have a well defined Axiom Virtual Machine that is not tied to Lisp.
>

As I understand it, this was also the goal of the specification of
FOAM for Aldor.

> | Are you aware that currently Aldor works within Axiom by compiling to
> | Lisp?
>
> Yes. But, eventually, I want to move away from Lisp.
>

I agree with that - or perhaps at least I agree that it is desirable
to identify a specific subset of Lisp that is required to support
Axiom and it's current algebra library. I am thinking here for example
about CCL as used in the last commercial release of Axiom.

> | How do you see FOAM being involved?
>
> I would like to see more documentation of FOAM and its semantics -- I
> know there is a documentation; but I'm not sure how current it is and
> whether it provides me with all I need.  I briefly discussed the
> documentation issue with Stephen when I was in London last July.
>

What do you need to investigate this issue more deeply? Are you
willing to devote some time to reviewing and expanding the FOAM
documentation?

Regards,
Bill Page.

_______________________________________________
Aldor-l mailing list
Aldor-l@...
http://aldor.org/mailman/listinfo/aldor-l_aldor.org

[Gabriel Dos Reis-2]

On Tue, 4 Sep 2007, Bill Page wrote:

| As an abstraction layer this makes very good sense. I believe that
| Aldor also provides an interpreter for this layer: "-gloop", right?
| But it is not clear to me how Axiom could make use of this.

I believe the Aldor compiler can re-load FOAM files and work with
them.  `All' OpenAxiom needs to do is to generate FOAMs for linking with
the Aldor object files.

[...]

| > | How do you see FOAM being involved?
| >
| > I would like to see more documentation of FOAM and its semantics -- I
| > know there is a documentation; but I'm not sure how current it is and
| > whether it provides me with all I need.  I briefly discussed the
| > documentation issue with Stephen when I was in London last July.
| >
|
| What do you need to investigate this issue more deeply?

As I said, it is not clear to me that the documentation currently
publically available contains all I need to know to generate (efficient)
FOAMs for Aldor compiler to use.  

| Are you willing to devote some time to reviewing and expanding the FOAM
| documentation?

Well, I suspect the documentation job is best done by someone who
underdtand FOAM much better than me.  When I (briefly) discussed this
with Stephen, he (jokingly) refered to legendary Unix manpages that
appear impenetrable to the novice, and crystal clear to the
experts who already know their contents :-)

-- Gaby

_______________________________________________
Aldor-l mailing list
Aldor-l@...
http://aldor.org/mailman/listinfo/aldor-l_aldor.org

[Bill Page-7]

On 9/4/07, Bill Page wrote:

> ...
> On 9/3/07, Gabriel Dos Reis wrote:
> > On Mon, 3 Sep 2007, Bill Page wrote:
> ...
> > | How do you see FOAM being involved?
> >
> > I would like to see more documentation of FOAM and its semantics --
> > I know there is a documentation; but I'm not sure how current it is
> > and whether it provides me with all I need.  I briefly discussed the
> > documentation issue with Stephen when I was in London last July.
> >
>
> What do you need to investigate this issue more deeply? Are you
> willing to devote some time to reviewing and expanding the FOAM
> documentation?
>

This reminds me to mention that the issue of FOAM and other Aldor
documentation was also raised at the recent Aldor Workshop. One thing
that came to light was some additional and more recent documentation
that was apparently translated from something originally in German. I
regret that I did not take specific note of the author or obtain a
photocopy of the document that I briefly examined at the workshop
(provided by Laurentiu Dragan) but I am sure that Laurentiu knows to
which document I am referring. I think that it might be very useful to
make this document available for download from the Aldor Wiki. If an
electronic format (e.g. dvi, pdf)  is not available than even a
scanned format would be useful.

Regards,
Bill Page.

_______________________________________________
Aldor-l mailing list
Aldor-l@...
http://aldor.org/mailman/listinfo/aldor-l_aldor.org

[Bill Page-7]

On 9/4/07, Gabriel Dos Reis wrote:

> On Tue, 4 Sep 2007, Bill Page wrote:
>
> | As an abstraction layer this makes very good sense. I believe that
> | Aldor also provides an interpreter for this layer: "-gloop", right?
> | But it is not clear to me how Axiom could make use of this.
>
> I believe the Aldor compiler can re-load FOAM files and work with
> them.  `All' OpenAxiom needs to do is to generate FOAMs for linking
> with the Aldor object files.
>

Having Spad generate FOAM code seems like a rather radical (and
interesting!) idea. How practical do you think it might be to make
such deep changes to Spad?

Regards,
Bill Page.

_______________________________________________
Aldor-l mailing list
Aldor-l@...
http://aldor.org/mailman/listinfo/aldor-l_aldor.org

[Gabriel Dos Reis-2]

On Tue, 4 Sep 2007, Bill Page wrote:

| On 9/4/07, Gabriel Dos Reis wrote:
| > On Tue, 4 Sep 2007, Bill Page wrote:
| >
| > | As an abstraction layer this makes very good sense. I believe that
| > | Aldor also provides an interpreter for this layer: "-gloop", right?
| > | But it is not clear to me how Axiom could make use of this.
| >
| > I believe the Aldor compiler can re-load FOAM files and work with
| > them.  `All' OpenAxiom needs to do is to generate FOAMs for linking
| > with the Aldor object files.
| >
|
| Having Spad generate FOAM code seems like a rather radical (and
| interesting!) idea. How practical do you think it might be to make
| such deep changes to Spad?

At the moment, I don't know.  
But, the existing compiler can already generates Lisp FOAM for some
codes.  I suspect that was the work of Peter Broadbery -- anyone to concur?

-- Gaby

_______________________________________________
Aldor-l mailing list
Aldor-l@...
http://aldor.org/mailman/listinfo/aldor-l_aldor.org

[Laurentiu Dragan]

Bill Page wrote:

> This reminds me to mention that the issue of FOAM and other Aldor
> documentation was also raised at the recent Aldor Workshop. One thing
> that came to light was some additional and more recent documentation
> that was apparently translated from something originally in German. I
> regret that I did not take specific note of the author or obtain a
> photocopy of the document that I briefly examined at the workshop
> (provided by Laurentiu Dragan) but I am sure that Laurentiu knows to
> which document I am referring. I think that it might be very useful to
> make this document available for download from the Aldor Wiki. If an
> electronic format (e.g. dvi, pdf)  is not available than even a
> scanned format would be useful.

Yes, I'll dig it up and put in on the web. The original FOAM
documentation is already published.

--
Laurentiu

_______________________________________________
Aldor-l mailing list
Aldor-l@...
http://aldor.org/mailman/listinfo/aldor-l_aldor.org

[Bill Page-7]

On 9/4/07, Laurentiu Dragan <ldragan@...> wrote:
> Bill Page wrote:
> > This reminds me to mention that the issue of FOAM and other Aldor
> > documentation was also raised at the recent Aldor Workshop. One thing
> > that came to light was some additional and more recent documentation
> > ...
>
> Yes, I'll dig it up and put in on the web. The original FOAM
> documentation is already published.
>

Excellent. Thanks, Laurentiu.

Regards,
Bill Page.

_______________________________________________
Aldor-l mailing list
Aldor-l@...
http://aldor.org/mailman/listinfo/aldor-l_aldor.org