RIS specification

larsgw

Sorry for bringing this up, it is not a particularly fun subject. I was wondering what the developer's stance on RIS specifications is? As far as I understand, there are two major official specifications, one from 2001 and one from 2011/2012. The new specification has many useful changes, like introducing tags for DOIs and the like. However, it also introduces a weird system where tag meanings depend on the entry type. I am pretty sure EndNote outputs RIS according to the new specification, but there might be some slight discrepancies.

Most other sources of RIS I have seen, like websites that export RIS, use a mix of both, either because of bad information when implementing (probably) or to (1) use DOI and other modern tags while (2) avoiding the complexities of the new specification. Interestingly, I took a quick look and Google Scholar seems to return valid "old" RIS so far. Because of all that, I understand that Zotero cannot really stick to a specification.

However, would it be reasonable for the translator code to (1) be more strict about RIS exports and (2) double-check that both (or either) specification is fully supported for imports? I could help with both. RIS exports are pretty good now but some specific things (like IS/M1 for issue numbers in magazine articles) are not entirely correct, and based on comments in the translator file I got the idea that that wasn't a known issue.

dstillman

I can't really speak to specifics, but the goal of Zotero's RIS support is real-world interoperability, which means doing the best we can with decades of EndNote versions that tended not to follow their own spec in novel ways, as well as lots of other software with varying implementations (in some cases just from the need to shoehorn modern fields into a decades-old format). The translator is full of conditional logic to try to make the best of a bad situation. The spec itself of limited use at this point.

If you have a question about a specific decision, you should ask it. There certainly might be decisions from, say, 2006 that are no longer warranted.

adamsmith

As a baseline, though, we rely on these, which I am pretty sure rely on the 2011 update of the specs:
https://github.com/aurimasv/translators/wiki/RIS-Tag-Map-(narrow)

larsgw

Ah okay, then it is probably just individual things such as M1 for magazine article issue numbers. I thought I read some comments in the translator file that labelled mappings based on that map as "EndNote hacks" but I might also have mixed things up. I will look for specific problems and get back to you (with a pull request, hopefully).