[Axiom-developer] Axiom bibliography - questions to consider

[C Y]

Hi everybody:

   Concerning the Axiom bibliography, Bob made the point that perhaps
we shouldn't worry about organization of the bibliography as a whole,
since its primary purpose is to serve as a repository for bib
references used in pamphlets.  I may have gotten a false impression
from earlier discussions, so before I go further maybe we should decide
definitely a couple of things:

a)  Do we want to include annotations as part of the bibliography?

   Personally, my feeling on this is that we should, and I believe it
is something Tim suggested we should do.  Virtually all of Axiom is in
the form of literate documents ala pamphlet files, which I approve of. 
In the one, specific case of the bibtex bibliography data however, I
believe a pamphlet (or at least a .bib.pamphlet file) is not the
correct way to make it "literate".  The document content (in whatever
eventual form) will be bibtex paper entries and our comments (if any)
about that paper.  This means that a pamphlet file would consist of
large amounts of "code" (the bibtex syntax) and then our annotations
about the paper in question, which are often likely to be fairly
minimal.  Given this, the thing that makes the most sense to me is to
not introduce the extra overhead of the pamphlet structure, but take
advantage of the fact that bibtex is already structured.  If we want to
"see" the bibliography as a whole (more on this later), we can include
it in a "bibliography" latex document which could provide the
"pamphlet" part of the system.  I like Bob's description of axiom.bib
as a database, as opposed to a document - I think that's a good way to
look at it.

   Next question - what do we want to "do" with the annotations?  Do we
want them to simply be attached to the bibliography as a note for
people who want to find notes about the paper?  Do we include the
annotation in each pamphlet's LaTeX output?  
   My thoughts on this, which I expect will not be universal, are 1) 
We should not include the annotations in pamphlet-level output by
default, to avoid possibly undesirable length of the document.  2) 
Rather we should have a link available to an entry in an ouput document
which does have the annotation available, if the user wants that.  3) a
build time option of including or not including annotations in pamphlet
files might be a good idea - this can be accomplished by changing which
bst file is used (presumably there is a way to do something like copy
axiom-plain.bst to axiom.bst prior to compiling the pamphlet files). 
The bst files will need to become literate documents.

a)  What do we want to do about the "complete bibliography as a volume"
idea?

   Number 2 above suggests one of the reasons I favor having a complete
document containing all bibliography entries - that complete document
will include the annotations for each entry.  Another reason I have
touched on earlier - with this document organized according to MSC2000
and PACS, it will be possible for a researcher to quickly access the
level of coverage Axiom has attained in a particular research subfield.
 It will also provide a mechanism for classifying Axiom pamphlets
according to subject, which I believe was of some interest earlier.  If
we view the axiom.bib file as a database, many interesting
transformations become possible - assembling mini-bibliographies for a
single subject, assembling a bibliography of pamphlets only, finding
all papers in all subjects from a particular journal, etc. etc. etc.
  Plus, if we are to publish Axiom someday as "volumes", we might be
able to provide a mechanism where the "global" bibliography volume
could group papers by subject with an indication of which pamphlets
they are cited by.  This might actually be possible with existing tools
(http://www.tug.org/pipermail/macostex-archives/2004-June/006647.html),
and if there is interest I will look into it further.  In this fashion,
we are more or less building a roadmap to the Axiom system via
mathematical subject.  It allows someone to ask the question "I wonder
if Axiom implements this idea from this paper?"  They can then check
the bibliography by subject looking for that paper, and if it exists
check the pamphlets referenced by it.
  Naturally, the latex document(s) needed for this must be generated at
real time based on the bibliography (and possibly the pamphlets
themselves).  Right now I'm using unix system tools.  Probably the way
to eventually do it would be to move the logic into Lisp and streamline
it, but while I'm not yet up to that I can at least get something
working which can eventually be transferred to Lisp. (We will probably
also want things like CL-PPCRE, which I'm not sure works under GCL -
particularly non ANSI GCL.)

   I think it will become very important, as we move forward, to have
some kind of system that is consistent, comprehensive, and agreed upon.
 While Axiom is now technically composed of literate documents, most of
those consist almost completely of source code.  The real work of
turning them into complete, properly referenced papers is still ahead
of us, so now is the time (IMHO) to put something in place which will
scale up as far as might be needed in the future.  Since this is a
decision and a convention that would have widespread results, I invite
comment and discussion.

Cheers,
Cliff

__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com