Create and maintain a "Findable" DOI for the Dandiset as a whole

[yarikoptic]

From slack by @satra:

https://help.zenodo.org/#versioning - i think
@Yarik
may have mentioned this in some thread. indeed zenodo has version specific and overall dandiset dois. i think we should adopt this and it will allow us to be clear about it just like zenodo is.
overall, there is a lot we can learn and implement like zenodo without reinventing it ourselves.

I totally agree and indeed I referenced to that feature of zenodo before. I think

• that at large we have it all implemented already since DLP goes to most released version
• we just need to mint 1 more DOI when dandiset is first published
  • point to dandiset DLP (not specific release)
  • we don't want to do that when created, dandiset without versions is too easy to remove, no promise of persistence etc
• upon each next release update metadata reflecting current state and new version

I have just checked on web UI for https://doi.datacite.org/dois/10.48324%2Fdandi.000029%2F0.210806.1511 it is possible to update already published (Findable) DOI with new URL and even metadata. So the procedure should be

• register a new DOI (check if could be just 10.48324/dandi.{dandiset.id} or need to have the suffix since this is prefix for all versions to come, thus might need smth like
  • 10.48324/dandi.{dandiset.id}/dandiset or
  • 10.48324/dandi.{dandiset.id}/dlp or
  • 10.48324/dandi.{dandiset.id}/latest.
• upon publishing a new version of dandiset, update this DOI record with new mutable metadata (title, authors, etc)
• we will reveal this DOI for the draft version (better this than none or fake) but have proper ones for published version

This would

• provide a solution to Stop injecting "fake" DOIs into draft dandisets #1709

[satra]

10.48324/dandi.{dandiset.id}

let's just do that if that works.

[yarikoptic]

@djarecka could you please check (e.g. try on datacite sandbox) if we could have both 10.48324/dandi.{dandiset.id} and 10.48324/dandi.{dandiset.id}/{version.id} DOIs, i.e. that one DOI is prefix for another?

[djarecka]

it's possible, at least as a "draft" on the 10.48324

I also tested on the testing api (10.80507) and I was able to publish both: https://doi.org/10.80507/DANDI.999999/1.0 and https://doi.org/10.80507/DANDI.999999

[yarikoptic]

Great, thank you @djarecka -- so let's proceed with 10.48324/dandi.{dandiset.id} for the overall DOI.

@waxlamp do you think your team could take on from here or should we formalize via design PR?

[waxlamp]

@waxlamp do you think your team could take on from here or should we formalize via design PR?

I have some plans to address all the DOI issues (including this one) so yes I / my team can take it from here.

[yarikoptic]

Since then we already started the

• Design document for the Zenodo like DOI per dandiset #2012

Could you have a look and provide feedback on either we should continue with it, or just join or even take over?