DataCite have confirmed verbally that they'll fix this.

How can I tell which DOIs are affected? I've got a couple that resolve just fine via the doi.org URI, but not with Zotero.

How can I tell which DOIs are affected? I've got a couple that resolve just fine via the doi.org URI, but not with Zotero.

@Jmuccigr : You can't tell from looking at the DOI: you need to know if the database where the corresponding metadata are storied is at DataCite or CrossRef. If the former, Zotero import won't work until the bug mentioned above is fixed at their end.

Hi, just a note to thank for the information on what's going wrong here and confirm that the fix will be hailed with much enthusiasm. The Pangloss Collection, an open archive of endangered languages, now has DOIs for all resources, with serious metadata (code available here), and seamless integration with Zotero would be super-cool.

(Here's one such DOI, if anyone wants to test: https://doi.org/10.24397/pangloss-0005709 )

How can I tell which DOIs are affected?

If you use https://search.datacite.org/works/{doi} in your browser and it shows a search result, then it's a datacite DOI. So for the ones above:
https://search.datacite.org/works/10.24397/pangloss-0005709

👍

Would it be at all possible to follow up DataCite? We'd really like to use Zenodo DOIs for import to Zotero... https://forums.zotero.org/discussion/comment/353170

DataCite Technical Director Martin Fenner kindly answered an e-mail query sent to him on this topic saying:
"this is an open bug that we will hopefully fix soon enough"
The message is from April 6th, 2020.
We too look forward very much to the fix.

@alexis-michaud thanks for posting this. Because we are dealing with a number of other important issues at DataCite right now, I am reluctant to give a specific timeline for a fix. But we will aim to resolve this in April.

@adam3smith: This seems to be causing a freeze in PDF metadata retrieval. We should obviously fix that in Zotero itself, but as a quick hack that we can push out right away, do you think we can change

Unfortunately "agency" is not yet available for all DOIs in our API, so would probably also not reliably work. We have some other fields that are probably unique (e.g. ClientId, ProviderId, state), but schema and agency are really the two fields that make the most sense.

Until the end of the year we will be adding many DOIs from other registration agencies to our index, and we have already added 8.5 million Crossref DOIs. As you can see in https://api.datacite.org/dois?client-id=crossref.citations, all these Crossref DOIs use schema http://datacite.org/schema/kernel-4, as we convert the Crossref metadata into DataCite XML. So thinking about it a bit more, agency is the best field to determine whether a DOI from our APIs is a DataCite DOI.

Obviously you fetch metadata via DOI content negotiation so should never run into the Crossref DOIs in our index. One question I have is why you don't directly ask for Citeproc JSON from DataCite in the DOI content negotiation, the implementation is much better than a few years ago.

To summarize, what is the best way to fix this? Should DataCite fix the inclusion of the schema "http://datacite.org/schema" as quickly as possible, but definitely before the end of April, or do you want to switch to another strategy to detect a DataCite DOI, e.g. via agency?

One question I have is why you don't directly ask for Citeproc JSON from DataCite in the DOI content negotiation, the implementation is much better than a few years ago.

I don't remember the details, but we experimented with this in January 2019 and found that we got significantly better import using the Datacite JSON

OK, I put in a temporary fix for Retrieve Metadata for PDF. The DOI Content Negotiation translator is now looking for the strings "agency": "DataCite" or "providerId": in addition to http://datacite.org/schema.

Unfortunately that wasn't sufficient, because at least for 10.7916/D8959HR1, the DataCite JSON translator then failed with data.subjects is undefined looking for tags, and if I put that in a conditional it failed with data.formats is undefined looking for medium. Others would have failed after that, so I just made most of the fields optional.

For reference, here's the request we're making:

curl -v -L -H "Accept: application/vnd.datacite.datacite+json, application/vnd.crossref.unixref+xml, application/vnd.citationstyles.csl+json" 'https://doi.org/10.7916%2Fd8959hr1'

When I tried just removing application/vnd.datacite.datacite+json from the Accept header, DataCite returned a 200 with Content-Type: application/vnd.crossref.unixref+xml and no content (which seems wrong). It only worked if I removed Crossref XML from Accept as well, in which case it returned CSL JSON.

But for now, with the above fixes, the DataCite JSON is working, at least for this DOI.