SPIRES, BibTeX, and the arXiv -- problems/missing features
Problem 1: the Spires site translator imports a paper's information via the bibtex info, but does not keep all fields; the preprint (arxiv) info is lost. A search on these forums reveals that other site translators also drop extra fields, even if it is coming from a bibtex import. These extra fields are quite important, and now pretty standard in the physics community. Even if, by default, this info would not be displayed in Zotero, it should at least be used when exporting (e.g. to create a bibtex entry for a latex document; it should have exactly the same info as grabbing it from Spires directly).
Problem 2: the Spires site translator never downloads any associated PDFs. Considering that all the journal links I've come across through Spires will itself import just fine with Zotero, could we have the option of using those translators directly from a Spires search page? In other words, rather than following links manually to journal websites (tedious if trying to save many papers from one search result, for instance), Zotero should give the option to attempt to automatically do this. At the very least, there should be the ability to save the pdf from the arxiv page, as Spires search results have direct links to the pdf on the arxiv (if available, and now just about a given except for old papers).
To summarize: please do not drop any extra bibtex fields on import (in general, not just for Spires), and incorporate getting pdfs for papers through the journal link or directly from the arxiv pdf link in the search results.
These small changes would make Zotero absolutely perfect for these essential physic papers websites (which are basically all anyone needs in high energy physics, for instance).
Problem 2: the Spires site translator never downloads any associated PDFs. Considering that all the journal links I've come across through Spires will itself import just fine with Zotero, could we have the option of using those translators directly from a Spires search page? In other words, rather than following links manually to journal websites (tedious if trying to save many papers from one search result, for instance), Zotero should give the option to attempt to automatically do this. At the very least, there should be the ability to save the pdf from the arxiv page, as Spires search results have direct links to the pdf on the arxiv (if available, and now just about a given except for old papers).
To summarize: please do not drop any extra bibtex fields on import (in general, not just for Spires), and incorporate getting pdfs for papers through the journal link or directly from the arxiv pdf link in the search results.
These small changes would make Zotero absolutely perfect for these essential physic papers websites (which are basically all anyone needs in high energy physics, for instance).
On fetching metadata from a separate site using another translator, I have never been able to get cross-site scripting to work in a translator. I think you run up against a security barrier that is built into Firefox on this one. There might be a way around it, but again it would be complicated, and probably fragile.
Another idea might be to have, for sites that give an entry in a fixed form already like Bibtex, this information as an attachment. Then you could have an option in exporting a library (or just entries) on what to do if you are exporting to a format which entries have an attachment in the same format. Again, this would also require some work to add.
As for the actual info that is relevant for Spires/arxiv, it looks like this:
eprint = "0911.0687",
archivePrefix = "arXiv",
primaryClass = "hep-ph"
while older papers don't include the "primaryClass" since that appears in the "eprint" field as "hep-ph/" before the number.
Maybe it makes sense to have these extra fields collapsible or easily hidden somehow, to keep things simple looking? In fact, I think that would be worthwhile for other entries as well, or some option like "collapse all blank fields".
Ok, I guess cross-site scripting is out for now, but grabbing a pdf (if available) from the direct arxiv link on a Spires entry would be very useful.
(The URL does not include the primaryClass, but the target page identifies it implicitly, judging from the breadcrumbs across the top of the page.)
Perhaps the translator for Spires could add this URL to the Zotero item, plus add the primaryClass as a Zotero tag (since it seems meant to work as a tag in the target system also)?
Also, as you point out, I think Zotero does use that identifier (primaryClass) in the arxiv site translator to generate a tag. It would be good for the Spires import to be consistent with that (presumably just copy the code of the arxiv translator).
I think the easiest solution right now that would incorporate everything would be to add those 3 Bibtex fields to Zotero, and grab the pdf from the arxiv link (constructed directly from the info as you say). The tag info should also be done as the arxiv importer does.
The arxiv site translator also has a bug, I think with swapping the actual journal and preprint info. I haven't looked at that in detail yet, and the best fix would rely on the above changes anyway (to use the real Bibtex fields above).
An older pre-print might appear as: "[arXiv:hep-th/0703196]" at the end of a bibliography entry, while a newer one would be "arXiv:0906.1273 [hep-th]" (again, at the very end of a reference).
So older papers do not have the primaryClass appear in the Bibtex entry from Spires, and it can be considered blank (this info is included in the eprint field). The newer ones do have this info, and while it is not needed directly for finding the paper on the arxiv, it is useful to quickly see what field (theory, gravity, etc.) the paper was listed as. I would put primary class on the same footing as eprint, as the "new" numbering is over 10 years old.
hep-th/0703196
0906.1273 [hep-th]
(with the archive field set to "arxiv")
For CSL-driven output, styles for target journals may need to be extended to do the right thing with the info, but that's easy to do with everything in place. BibTeX export is out of my league, but deriving BibTeX entries for export from the content of the Zotero database is definitely the right thing to do. If it's broken it needs to be fixed, but the Z infrastructure is reliable and robust and it will be simpler in the long run to rely on it.
(EDIT: amended sample field content to remove arxiv: prefix.)
For the translator and CSL style stuff, I'm kind of loaded up with work at the moment, but can take a look when things slow down a bit. Is there a style or a set of styles that are commonly used in the field -- or the name of a leading journal or two (so I can pick it out of the Zotero CSL styles repository)?
(EDIT: Sorry, missed the link to the BST file above. Is IEEE Transactions different from IEEE?)
However, either Bibtex or a direct Latex (\bibitem) form is what I think everybody uses. Since Zotero does Bibtex export, getting that to match the Spires entry would be the most useful. Getting a journal style export would probably not be used in writing a paper (papers are submitted as a latex document), but maybe for some other uses...?