mEDRA: Broken, not detecting

[AbeJellinek]

Site layout has changed.

[abaevbog]

I briefly looked into this out of curiosity. I was wondering if there is a particular reason why mEDRA translator tried (before being disabled) to manually parse the <td> in DOM when there is a DOI readily available? Would it not be easier and more reliable to just grab the DOI and use the search translator on that? I suppose this is a general question of which approach should have the priority.

[AbeJellinek]

I haven't looked at the differences here in detail, but we usually try to use the site's own metadata for reliability/decentralization/consistency unless it's notably worse or less complete than the DOI metadata.

[AbeJellinek]

Oh, well, in this case mEDRA is the DOI agency, so the DOI resolves to them. Right now we handle mEDRA DOIs generically by parsing the CSL JSON returned by DOI.org's resolver, but if possible we should have this parse whatever their native format is so we can avoid a lossy conversion. It looks like they have an API that returns a format called ONIX, so we should parse that, probably in a separate import translator like we have for Crossref Unixref XML.

[abaevbog]

I haven't looked at the differences here in detail, but we usually try to use the site's own metadata for reliability/decentralization/consistency unless it's notably worse or less complete than the DOI metadata.

Gotcha! I'll keep that in mind. My initial gut instinct was to think that relying on DOI would be better, since it's less likely to be affected by the website changes.

It looks like they have an API that returns a format called ONIX, so we should parse that, probably in a separate import translator like we have for Crossref Unixref XML.

Right, that makes sense. Thank you for the explanation! All of this will go into my larger attempt to somehow document the translator patterns we currently have. After I went through existing translators to put together an outline, I figured it would be useful to go through some of the outstanding issues, which brought me here! This mEDRA translator issue should be a good practice.