[Feature]: Better reading of RO-Crate metadata files

[kmexter]

Detailed Description

I recently had a meeting with @huberrob and colleagues, in the context of the FAIR-EASE project, to discuss our use of the FUJI fairness checker on our metadata "records" which are actually ro-crate metadata json files. I ran the some files (see list below) through the checker and from a look at the report I got back, they are perhaps not quite read properly by FUJI. I mentioned this at the meeting and Robert asked me to provide these ro-crate examples for FUJI to check out

https://github.com/arms-mbon/data_release_001/blob/main/ro-crate-metadata.json
https://github.com/arms-mbon/analysis_release_001/blob/main/ro-crate-metadata.json
https://github.com/arms-mbon/code_release_001/blob/main/ro-crate-metadata.json

We made quite some effort to add lots of machine-understandable metadata to these rocrates, so I think they should give a good assessment. So we have identifiers, download URLs, licence, links to related entities, etc.
FYI The domain of these metadata is marine biology.

I know that there are 2 ways that I can specify the URL in the tool in order to get the best reporting, i.e.
https://raw.githubusercontent.com/arms-mbon/data_release_001/main/ro-crate-metadata.json vs https://github.com/arms-mbon/data_release_001/blob/main/ro-crate-metadata.json
It would be nice to have a recommendation as to which is the better approach (since the first one gave a better result, I am guessing it is that one)

Context

If you can modify FUJI to work better on ro-crates, we could trust the results better. We would also value any feedback from you on our ro-crates!

Possible Implementation

[huberrob]

Just a quick first comment: only the https://raw.githubusercontent.com style URLs are pointing to JSON content, URLs without the 'raw' subdomain are delivering HTML content in which JSON is displayed in a textarea tag, therefore F-UJI would not find it.

[mpo-vliz]

I know that there are 2 ways that I can specify the URL in the tool in order to get the best reporting, i.e.
https://raw.githubusercontent.com/arms-mbon/data_release_001/main/ro-crate-metadata.json vs https://github.com/arms-mbon/data_release_001/blob/main/ro-crate-metadata.json
It would be nice to have a recommendation as to which is the better approach (since the first one gave a better result, I am guessing it is that one)

There is a third way, well rather the imho "correct" way

My own recommendation is not to use either of those refs — we need to stop using github-based uri in these case, mostly because they are accidental to our current located host, and entirely not core to these data 'identity', definitely not a dependency we want to be tied to. We did an effort to organize the data.arms-mbon.org domain -- and so the correct way to publicly refer to that ro-crate is https://data.arms-mbon.org/data_release_001/

From there possible redirects + a finel html-embedded (link;rel=descibedby) fair-signposting-conform, should lead one to the metadata,json file in question.

[cedricdcc]

quick sidenote. The page that you are describing here https://data.arms-mbon.org/data_release_001/ is a collection of versions of a rocrate. These don't have LOD yet but will have that in the future. The page that has embedded fair-signposting is https://data.arms-mbon.org/data_release_001/latest/ which is the latest version of the data_release_001 rocrate.

[huberrob]

True, at least F-UJI would follow the describedby links to locate the metadata.

[huberrob]

Maybe you could replace the fileFormat values with correct mime types?
Not sure how this is handles in RO-Ctate namespace butaAs far as I know fileFormat is replaced in schema.org by encodingFormat

[mpo-vliz]

I see the rocrate context has provisioning for both encodingFormat and fileFormat

And the spec recommendation is to use encodingFormat to hold the mime types
https://www.researchobject.org/ro-crate/specification/1.2-DRAFT/data-entities.html#encoding-file-paths

[kmexter]

FYI We are upgrading our ro-crates that I gave as examples - will ping you when we are doing (will not be until next year).
From comments on #543 I wonder if also here there is an issue with the fact that a ro-crate can describe multiple datasets (so effectively has multiple download URLs). FYI however, if a ro-crate is filled properly then it is, in fact, possible to link each metadata element to named data elements. An alternative approach is to consider the entire ro-crate as being the jsonld metadata record of a single dataset, albeit one with many elements.
Anyway, thoughts? Should I continue feeding this issue, or is this enough?

[huberrob]

OK sounds great, it should be no problem to pack several Datasets in one ro-crate, but F-UJI will evaluate the root(the crate which is of type CreativeWork) not each individual Dataset. Looking forward to see these examples!

[kmexter]

so it will fail all tests related to metadata provided for the dataset itself, right?