504 Gateway timeout on specific DOIs pointing to pdfs

[rduyme]

Hi,

I tested a few times, a few different days. the DOI : 10.19276/plinius.2025.01.014 on

https://tests.ostrails.eu/community-tests/

• community_funding_information_registered => 504 Gateway timeout
• community_metadata_includes_author_affiliation => 504 Gateway timeout
• community_license_information => 504 Gateway timeout
• community_open_access_publication => response OK

Same days no problem with other DOIs.

Note : 10.19276/plinius.2025.01.014 is a Medra DOI (https://doi.org/doiRA/10.19276/plinius.2025.01.014) , but other Medra are ok ex: 10.1392/roma081203

[rduyme]

Still 504 issue this morning on this specific case.

The rest of the Champion platform is running fine.

[markwilkinson]

That's a wierd one. the DOI resolves to a PDF rather than a landing page, following all kinds of paths that attempt to gather metadata. That PDF is then sent to a variety of parsers, which are very unhappy consuming a PDF. Eventually they die.

I can solve this by not putting your test input through my normal metadata harvesting pipeline, and using ONLY the specific DOI API calls that you have specified.

Tell me if that is what you prefer.

[markwilkinson]

Further to that, note that Allyson may approach you today or tomorrow to request exactly the opposite of what I suggest above! So... it may be necessary to find some way to catch these "edge cases" before they are sent to the parsers. Parsers like Tika should be able to deal with PDFs, and there could be useful metadata embedded inside the PDF, but I can catch it before it is sent to other parsers that don't know what to do with a PDF. That's probably the safer route.

[rduyme]

I think he byte check for pdf works ok : https://stackoverflow.com/questions/6186980/determine-if-a-byte-is-a-pdf-file.
It seems TIKA has feature to detect. https://www.baeldung.com/java-check-file-pdf
I suppose you can exclude the pdf parsing, it is a very very rare case I see ( at least in the 50 000 pubs corpus, I deal with ).

In the end, DOIs pointing to pdf, is a bad practice for DOIs https://support.datacite.org/docs/landing-pages#dois-should-resolve-to-a-landing-page-not-directly-to-the-content and also for FAIR I suppose. It should raise issues in one of the Core FAIR tests ( but which one F3 ? :-) )