Joss: Link JOSS and AAS DOIs to Zenodo

[dbouquin]

For software reviewed by JOSS in collaboration with AAS, there needs to be a step in the process where both the JOSS and AAS identifiers are added to the software's archival record

Why this would be helpful: PRISM example
(ApJ paper; JOSS paper; Zenodo record)

• PRISM's Zenodo record points to arXiv, not the ApJ DOI -- Adding the ApJ DOI directly would prevent the need to rely on access to ApJ via arXiv's link to ADS.
  
  
  
  • If the arXiv link wasn't added to the Zenodo record there wouldn't be a bridge from the code archive itself to the ApJ paper at all.
  
  
• PRISM's Zenodo record also does not contain a link to the JOSS paper -- instead it bundles a version of the JOSS paper in with the code, but there's no metadata to indicate that it's there (also only contains a .md and a .bib but no PDF).
  
  
  
  • Neither the .md or .bib files contains any reference to the JOSS DOI, and since the Zenodo record itself also does not reference the JOSS paper, there is no link from the code archive itself to JOSS where the code review could be found (there's also no date on either the .md or .bib files).
  
  
  • Neither the .md or .bib files contains a reference to PRISM's Zenodo DOI. So when you to download the archive bundle from Zenodo there is no indication that it was ever on Zenodo, which would be the place to find out if there were other versions archived or other papers/code reviews associated with those versions.
  
  

If PRISM's Zenodo record just referenced both the ApJ and JOSS DOIs as related identifiers this would all be stored in the DOI system and you wouldn't need to rely on software authors knowing to build the right bridges.

[labarba]

If I got this right, we (JOSS editors) should be asking the authors to add the JOSS paper DOI and the ApJ paper DOI to the "Related identifiers" in the Zenodo archive's metadata. Correct?

JOSS already includes the DOI of the software archive as a <rel:related_item> of the article metadata. Maybe for articles in collaboration with AAS it can also include the DOI of the ApJ paper?

[dbouquin]

You got it. If JOSS authors added the JOSS paper DOI and the ApJ DOI as related identifiers in their Zenodo record, the semantic relationships between all three objects would be unambiguous and direct (and way less prone to link rot).

If you were to add a <rel:related_item> to the JOSS article metadata for the ApJ paper that would be excellent. Right now it's embedded in the PDF but a direct bridge from record to record is better.

[arfon]

JOSS already includes the DOI of the software archive as a <rel:related_item> of the article metadata. Maybe for articles in collaboration with AAS it can also include the DOI of the ApJ paper?

This item should be relatively easy for us to do.

If I got this right, we (JOSS editors) should be asking the authors to add the JOSS paper DOI and the ApJ paper DOI to the "Related identifiers" in the Zenodo archive's metadata. Correct?

This one is a little harder as we're going to have to rely upon our authors to get this right.

This reminds me of an idea we had some time ago of using Whedon to create the archive of the software on behalf of the JOSS authors. That way we could control that record and add the necessary metadata we needed for cross-linking like this.

[augustfly]

oh, here is the corresponding issue for ensuring all object-object bonds are created. I should have known @dbouquin wouldn't miss this!

From my experience (it is in our documentation) the go-back and add relations step is a hard one to get authors to do without followup. In part I think it is because of the asynchronous nature of backfilling metadata. They often add the arXiv link if they have one. I think that the logic of the "why do this this way" escapes them without someone at least asking them to do it.

Lastly, IOP (our publishing platform) doesn't know how to pass CrossRef relations (yet); we are working on them doing this in their future content model and could move the JOSS article in our metadata from a direct citation to a references relation if that distinction were important.