Add Zenodo references/auto-publish?

[fmigneault]

Description

In the context of a scientific paper, I recently produced CRIM Hirondelle: https://zenodo.org/records/18946328
It itself references birdhouse-deploy: https://zenodo.org/records/18856887

Since I can't push our private code, and that Zenodo requires at least one file upload, I used the JSON of the current node-registry filtered to only our instance, and added a $schema reference to the node-registry's schema for increased diffusion of the source.

During that process, I though, why not make all the nodes auto-publish their Zenodo references, just like birdhouse-deploy does on each release? Notably, the node-registry's CI detects when to update the definitions from a version/metadata change from the node's server, so it could publish the Zenodo update at the same time and make sure all links are nicely integrated/cross-referenced between projects.

To support it, we would only need an extra link, which could conveniently be the official IANA Link Relation cite-as. And it would skip if the updated instance does not provide it, so it doesn't force anyone to provide it if they don't want to.

Relevant Files

Since each note would have their own "release", we cannot rely on a global and auto-integrated Zenodo config:

• https://github.com/bird-house/birdhouse-deploy/blob/master/.zenodo.json

However, we could have the operation within the CI, as such:

• https://github.com/crim-ca/weaver/blob/master/.github/workflows/zenodo.yml

And maybe using a matrix job or similar to define the doi parameter (https://github.com/crim-ca/weaver/blob/4b433c1b5ffe1a8aae9673263cf6ce60e1cfe487/.github/workflows/zenodo.yml#L50-L55) on every node's reference with a conditional if when updates apply.

---

❓ Does your instances already have Zenodo DOIs?
❓ Thoughts about the feature, is it of interest?

[fmigneault]

@QSparks FYI - couldn't add you to the "assignees"

[huard]

I like the idea of having a citable object for a running node.

I think the requirements for this to be useful to researchers are:

• It's easy to find the DOI of the running node (e.g. /cite);
• There's a table with dates and DOIs so you can go back and get the DOI for the node for a given date.;
• The content of the DOI includes the version of the individual components, and their DOIs if they exist.

[fmigneault]

@huard
I like the idea of the table per node. I think it could be displayed in the repo's README and also the MarbleClimate website, though I would limit it to the "all versions DOI" to keep it concise. If needed, the specific versions/updates would be listed on Zenodo itself.

Good point about including the version and components. It could also include the services response. All 4 JSON (including the schema here of the node itself) could be packaged together in the Zenodo upload.