Re: [Axiom-developer] Axiom bibliography

[C Y]

--- Bob McElrath <address@hidden> wrote:

> C Y address@hidden wrote:
> > JabRef would be far and away my
> > preference - it is the leading open source solution that I am
> > aware of, and I'm hoping its Groups feature could be used somehow
> > to manage things:
> > 
> > http://jabref.sourceforge.net/
> >
> http://jabref.sourceforge.net/images/Jabref-ScreenShot-MainWindow.png
> > 
> > but if we do use such a tool we have the problem of its being
> > designed to work with bibtex, not bibtex in pamphlet :-/.
> 
> I'm working on a system for physicists which will require an
> extensive reference database (at least all of the arXiv plus all of
> SPIRES) and furthermore, I want wiki pages to actually link to the
> bibliography.
> 
> For physics anyway, search engines already exist to find papers and
> even entire bibtex entries (via arXiv, SPIRES, and NASA ADS).  For
> math see the Front here at UC-Davis which also provides bibtex:
> http://front.math.ucdavis.edu/  AND these systems all provide a
> concise identifier for each paper.  (e.g. math.AC/0604301)  Historic
> papers are often more of a problem.

Isn't that a bit different from the problem of how to handle axiom.bib?
 Web interfacing and updating of it (with convenience links to online
archives) would definitely be a Good Thing but for the pamphlets we
need a bib file.  And for us non-web-editing fogies we need some
desktop tools to handle it as well.  Getting all of these worlds to
speak to each other cleanly seems to be as yet something of an unsolved
problem.

> I have discussed this project a bit with Bill privately, but would
> welcome more widespread development in this direction.

I think I haven't wrapped my brain all the way around this one, but I'm
going to keep trying to do so because it is extremely important to a
literate programming project such as this one.

> Right now I have an experimental Plone site (because of the existence
> of the Bibliography Archetype -- same as Axiom Portal:
>     http://portal.axiom-developer.org/bibliography_view
> ).  Maybe I have missed it but I do not see a way to search the
> bibliography.  Furthermore when viewing a bibliography entry it does
> no give you a concise identifier for each paper that you can \cite{},
> unless you go export to bibtex, then examine the output.  

Agreed.  OK, so you are looking at a web interface that parses the
axiom.bib file, allows searching of it, and provides convenient
citing-via-web abilities as well as links to actual papers (when
available?)  Presumably entries could be added too, although the manner
in which they are added needs careful thought.

> I would like to add search and \cite{} capability from wiki pages. 
> This would allow discussion of papers, and even back-references on
> the wiki! It would be super-sweet if the search could be done in a
> sidebar, next to the edit window, using AJAX, so that you could
> retrieve the identifier for \cite{} without having to open a new
> browser window.

OK, I think I'm beginning to follow you.  Yes, that would indeed be
nice!
 
> > 1.  Create and maintain a central bibliography, starting with the
> > Nelson Beebe axiom.bib file as a beginning.  Pamphlet authors would
> > be encouraged to use the citations already present if they want to
> > cite an existing paper.
> > 
> > 2.  Each new pamphlet will undoubtedly cite papers and other works
> > not included in the main axiom.bib file.  They will most likely
> > either append these new entries on to the end of axiom.bib or
> > maintain a separate file.  When a new pamphlet is submitted to
> > Axiom, part of the inclusion process will be incorporating the new
> > bibliography entries into the main axiom.bib file.  This can be
> > done either via a diff (preferably of an altered axiom.bib which
> > has been sorted by whatever criteria we decide to organize the
> > main file) or the inclusion of a bib file which contains all the
> > new references.  In the latter case, someone with the correct
> > tools can merge the new references into the main file.
> 
> I have been thinking to write a back-end which will interrogate other
> search engines for bibtex, if someone enters something unknown.  For
> instance, SPIRES references are easy to recognize by syntax (e.g.
> McElrath:2005bp) same for arXiv (e.g. hep-ph/0506151) so the wiki
> system can know which database to import from if it doesn't have the
> bibliography entry already.

Two concerns there.  The first one is how we ensure quality and
avoidance of duplications (suppose a slightly different bibtex entry
already exists for that paper - how do we spot it?) but I think the
online archives are reasonably good quality so I'm not too worried. 
The second is copyright - if we just snatch bibtex references for
papers off the various sites and incorporate them into axiom.bib, is
that OK?  I can't imagine why it would be a problem since the whole
idea of bibtex entries is to use them in papers but the question needs
to be asked.

> > 3.  We need some way to identify categories of paper within the
> > main bib file - maybe some kind of agreed upon keywords for each
> > category of paper.  This will evolve somewhat but most major
> > categories should be agreed upon in advance.  This will allow
> > someone looking for a particular paper to cite to quickly identify
> > the relevant subset of the bib file.  This will require jabref or
> > some similar tool to handle properly, but that is going to be
> > unavoidable in such a large file.
> 
> Rather than explicitly identify categories, why not just use
> backreferences?  e.g. a wiki page called CommutativeRings has a
> number of references.  If you viewed one of those entires in the
> bibliography, it would show "pages that cite this paper" including
> your CommutativeRings.

That would work for the web, but what about "stand alone" authors
working in (say) emacs?  Or what about a case where someone says "how
extensive is Axiom's research on topic X" and they then want to check
the bibliography for all papers pertaining to X? Or "have any papers on
topic Y from the last Z years been considered by Axiom?" 

> > 4.  If things get too large, we might have to consider "group"
> > based bib files - I don't know what the size limits are for bibtex
> > but it must have some.
> 
> I say one huge database.  I know of no size limits.  If one wants to
> process with latex one could always extract the subset of bibtex
> entries from the database.

OK.  Will we adopt the policy that all entries in the axiom.bib file
must be cited by some pamphlet?  Otherwise we will have references
which will never appear in any pamphlet but still exist in the
database.  There are some cases in the past where I have wanted a
reference in my bibliography even though I didn't cite it in a
particular paper because it could be relevant to later work - is this
bad practice?

> > 5.  IIRC there is some way to process bib files as TeX documents? 
> > If so we can make the bibliography file available on the wiki like
> > any regular pamphlet file.  I'll have to check that.
> 
> What do you mean here?  Have you seen the Axiom Portal bibliography
> that I'm talking about?

Yes - I was thinking about whether we want to make an Axiom volume out
of the comprehensive axiom.bib contents or not. Probably not, so never
mind. 

> If you're interested in working on this, please do!  For the near
> term I'm trying to get an XHTML+XML plone site working, and 
> developing a new wiki syntax that is faux-latex and generates
> XHTML+MathML.  I have not started on the bibliography stuff yet, but
> I will need some kind of interface to query the bibliography.  
> e.g. "give me a formatted citation or URL for the identifier
> McElrath:2005bp".  If you want to write that, or look into it, 
> please do!

Unfortunately, I'm thinking more about axiom.bib itself than the web
interface at the moment - I do my editing offline so a web environment
is less useful in the short term.  Eventually all the various
environments, text, jabref, and web alike, will have to funnel down
into one axiom.bib file, so I was more concerned as to what that file
should look like.

You raise a very important point - for a large number of papers the
identifier (and probably bibtex file) should conform to what is
commonly used online. I guess I will need to search for all my
references in the common online archives.  citeseer I'm familiar with,
and xxx.lanl to some extent, but we should have a policy of "check for
auto-generated bibtex entries from these archives before creating your
own."  We should also decide on one convention for identifiers (and
quality standards) on papers where we need to create our own bibtex
entry - the pamphlet files themselves being an obvious example.

Cheers,
CY

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