Re: [Axiom-developer] Community

[C Y]

Tim, I think part of the difficulty here is that there is no universal
consensus about direction yet, so the community is still testing
various possibilities.

For example, I am currently looking into ASDF with the idea of teaching
it how to load pamphlet files.  (A tad harder than I had hoped, but I
think it can be made to work.)  If that is possible, and cl-web grows
all the necessary features, it should be possible to have a lisp-only
asdf based build mechanism for all parts of Axiom defined in Lisp, Boot
or SPAD. E.g. the boot/shoe compiler could be described by something
like (I suppose the files are wrong):

(defsystem #:boot
  :name "BOOT Language Compiler"
  :author "The Axiom Team"
  :version "1.0"
  :components
  ((:module "BOOT-BOOTSTRAP"
            :pathname #.(make-pathname :directory '(:relative "src"
"boot"))
            :components
            ((:cl-pamphlet "ptyout" :chunk "ptyout.clisp")
             (:cl-pamphlet "btincl2" :chunk "btincl2.clisp")
             (:cl-pamphlet "typrops" :chunk "typrops.clisp")
             (:cl-pamphlet "btpile2" :chunk "btpile2.clisp")
             (:cl-pamphlet "btscan2" :chunk "btscan2.clisp")
             (:cl-pamphlet "typars" :chunk "typars.clisp")
             (:cl-pamphlet "tytree1" :chunk "tytree1.clisp")))
    (:module "BOOT"
            :pathname #.(make-pathname :directory '(:relative "src"
"boot"))
            :components
            ((:boot-pamphlet "ptyout" :chunk "*")
             (:boot-pamphlet "btincl2" :chunk "*")
             (:boot-pamphlet "typrops" :chunk "*")
             (:boot-pamphlet "btpile2" :chunk "*")
             (:boot-pamphlet "btscan2" :chunk "*")
             (:boot-pamphlet "typars" :chunk "*")
             (:boot-pamphlet "tytree1" :chunk "*"))))

And so on for the rest of Axiom.  This would mean traditional asdf
commands could be used to build parts or all of Axiom.  (Maxima uses
defsystem in a similar fashion, although of course they don't have to
worry about pamphlets.)

Even if I do get this working however, the community may not care for
it.  Gaby has put a great deal of work into autoconf (which IS in
essence a competing solution, and has the considerable merit of already
working), and the above system would almost certainly require a
properly working ANSI lisp.  It would replace those parts of the
current build systems that use Makefiles to build lisp, boot or spad. 
The dumping of images is another sticky point I haven't gotten to yet. 
Some folks seem inclined to avoid Lisp as much as possible.

So you see the difficulty.  Even if I succeed preparing cl-web+asdf
extensions as a working alternative, the community may decide against
it.  But it is part of where I personally would like to see Axiom head,
so I'm doing what I can to make it work.

There are other issues.  bookvol5 style code grouping vs. "conference
proceeding" style volumes on topics vs. *other-favorite-method*, for
example.  We have had very interesting discussions on these issues but
I don't really see a consensus forming yet.  Personally I want to try
out some mechanisms to see how well they work.

Bibliography issues need to be resolved, and I'm still working on how
best to handle that issue.

Cheers,
CY


      
____________________________________________________________________________________
Park yourself in front of a world of choices in alternative vehicles. Visit the 
Yahoo! Auto Green Center.
http://autos.yahoo.com/green_center/