Need to check bibtex volume for crossref

[kmccurley]

One article cc2-3-73 submitted bibtex with

@article{EC:CanGold99,                                                                                                           
        author = {R. Canetti and S. Goldwasser},                                                                                 
        title = {An efficient threshold public key cryptosystem secure against adaptive chosen ciphertext attack},               
        journal = {J. Stern, editor, EUROCRYPT’99},                                                                              
        volume = {1592 of LNCS, pages 90–106,Springer, Heidelberg},                                                              
        Month = {May},                                                                                                           
        Year = {1999},                                                                                                           
        doi = {10.5555/1756123.1756134}                                                                                          
}

This is wrong in many ways, and apparently our bibtex parser did not catch things.

1. it should be @inproceedings, not @Article
2. the journal field is wrong.
3. the volume field is wrong.
4. they didn't put the pages into the pages field.
   This was only flagged because when we submit crossref references, it checks that the volume field can only be at most 32 characters.

This entry shows that we need to have a lot better checks on things. Distinguishing @article from @inproceedings is hard enough, but other fields need to at least be checked for length requirements.

[kmccurley]

It's not really feasible to check that a bibtex entry is properly constructed, because as long as it masquerades as something reasonable, the bibtex parser will format it. I found a bunch of examples where authors are confused about the semantics of @Article, @inproceedings, @incollection, @TechReport, and @misc. Examples include where a CoRR article was labeled as @Article when in fact it's a preprint so it should be @misc or @TechReport. Another was a FOCS paper listed as @misc in cc2-3-33 because the author was too lazy to look it up. They also omitted URLs for eprint papers and use @Article for a CCS paper (with no journal entry). There are a few heuristics we could apply to flag these, but they are at risk to produce too many false positives. I think the only feasible solution at this time is to check the entries to make sure they fit inside the crossref, and either flag them here or truncate them when we produce the actual crossref. A few heuristics to produce warnings are:

1. journal field should not be too long.
2. @misc or @online should not have a DOI. This happened in cc2-2-75.
3. if "eprint" appears in a journal field or booktitle field, this is probably mislabeled. This happened in cc2-2-6.
4. if STOC or FOCS appears in the journal field, then it's probably not @Article. This was in cc2-3-73. They also had CRYPTO papers labeled as @Article.

[kmccurley]

The one that flagged this first had the conference title shoved into the volume field. We should check that to see if it's numeric, and flag it as a possible error otherwise. It might still be ok, since IRE Transactions on Electronic Computers has volume numbers like "EC-10" (see https://ieeexplore.ieee.org/document/5407919). Note that DBLP labels CoRR articles as @Article. See https://dblp.org/rec/journals/corr/MilutinovicHWK17.html?view=bibtex for example.