Re: [Axiom-developer] Axiom bibliography

[C Y]

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

> C Y address@hidden wrote:
> > > The contents of the \cite{} should identify the
> > > paper, rather than the local bibliography ('axiombib') to look
> > > in. 
> > 
> > Well... I would prefer it that cite references axiombib first and
> > gets the URL from that, and if it is not present there goes 
> > elsewhere.
> 
> The local database would still be kept, and periodically updated from
> external database queries.  Sorry for the confusion -- I always
> intended that.  Solely external lookups would be slow and error 
> prone.  So all lookups would look in the local database first, then
> go to arXiv/ADS/etc if the requested entry isn't already in the 
> local database.

Cool!  In that case, it gets my vote!

> I was thinking this afternoon that the footnote could be "This
> citation is not in the database.  Reload to see it."  Where in
> between the server would be interrogating arXiv (etc) servers. 
> Or "This citation could not be found in any databases.  Click here
> to enter it manually."

I'm not sure what the limits of web browser interfacing would be, but I
imagine something like the travelocity site would be nice:

1.  If search local is successful, return immediately.
2.  If not, flash a "Reference not in local database, searching online
archives..." page while the search is going on.  Some kind of running
progress line like "Searching arXiv..."  "Searching ADS..."  etc. would
be nice but probably even more work.
3.  If search succeeds, return "Reference located, added to Axiom
database."  If it fails, return "Reference not found, please enter."

I take it we aren't going to worry too much about a particular paper
being "on topic?"  I guess in Axiom's case it would be hard to say with
certainty any particular paper couldn't be relevant.

I'm intrigued by the idea of a DOI number - something like that could
potentially be quite useful - but do arXiv papers, for example, have
DOI numbers assigned?  Also, I think DOI numbers might be inconvenient
and unintuitive for a \cite{} entry - I guess the best thing to do is
record a DOI number in the bibtex file if it exists, but don't use it
for the identity tag.

Maybe we should establish a pattern for choosing the tag for a paper. 
Something like the following:

1.  If available, use the tag from arXiv, ADS, or other database since
this will match others using those databases.  In the case of
duplication of a paper in differrent databases, we will need to
establish a priority.  I'm not familiar with all the options - can
someone suggest a logical a->b->c search priority for the various
databases out there?  Alternatively, we might use OAI conformance or
some other such criteria.

2.  If that the paper does not exist in any archive, we must have some
rule for creating a tag name that makes sense.  This will take a little
thought.  CiteSeer seems to like the work done by these folks:  Open
Archives Initiative http://www.openarchives.org/  I would suggest we
adopt whatever open conventions have been developed - maybe even create
a "Bibliography Howto" for inclusion in Axiom?  Also, if the OAI is
doing quality work that is relevant to our needs we can define priority
for database searches based on an archive's use of OAI-PMH standards -
check the OAI compliant databases first, then check other ones.

> > What about papers that are not listed in any of the major online
> > references?  I think we have to allow for an updated axiom.bib,
> > but I agree automating as much of it as possible through
> > connections with servers is a Good Thing.
> 
> Of course manual entry would be possible, but I hope that it would be
> rare.

Agreed :-).

> > > If you want to process a pamphlet in latex it's easy enough to
> > > extract the bibtex entries used and export them.
> > 
> > Maybe, but what if an online archive changes their format and
> > decides not to allow simple automation of bibtex entry downloads?
> > How will we store annotations?  I think a local solution that is
> > refreshed and updated periodically will be the best compromise -
> > we always have something that works locally (also for offline
> > editing, which does happen), but it is kept current with 
> > developments elsewhere.
> 
> Well that's an unfortunate possibility that would have to be dealt
> with. Fortunately, the number of databases is large.  But yes, the
> loss of a database would suck.

Is there some sort of automated way to check URLs included in bibtex
entries, in order to spot dead links?

> If that happened though we'd still have the local database of
> references *used* on the site.  We just wouldn't be able to obtain
> new ones.

OK, I get it now.  Sounds great!

Cheers,
CY

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