ResearchObject / ro-crate

Research Object Crate
https://w3id.org/ro/crate/
Apache License 2.0
79 stars 34 forks source link

Soften Profile inheritence rules #262

Closed stain closed 1 month ago

stain commented 1 year ago

In some cases, such as Workflow Run Crate there are multiple profiles that are related with inheritance.

There are also less strict inheritance profiles such as https://trefx.uk/trusted-wfrun-crate/0.3/

This section now makes it softer how to list multiple profiles, they are suggestions rather than enforcing any hierarchy.

ptsefton commented 11 months ago

Didn't we decide to remove all this from 1.2? @stain

ptsefton commented 9 months ago

This needs to be moved to a different document, right?

stain commented 4 months ago

Remove inheritance, at best "referenced/suggested" profiles.

stain commented 2 months ago

Decision 2024-04-11: Not do formal inheritance system at RO-Crate level yet before we have tried out how it works with actual implementations (e.g. SHACL, Crate-O modes) -- then facilitate them.

stain commented 2 months ago

I have finally managed to tidy up the profile section.

Changes:

stain commented 2 months ago

Suggestion in call 2024-04-25 @CaroleGoble was to also look at Delegation pattern:

In a sense we use delegation pattern in https://trefx.uk/5s-crate/0.4/#referencing-a-workflow-crate as instead of subclassing Workflow RO-Crate, it references the Workflow RO-Crate by URL -- one crate to another. I've documented this pattern in #296.

This would make the resulting 5s-crate as Workflow Run Crate not conform to any strict inheritance rules, as it would not have the usual ComputationalWorkflow etc. and so should not itself declare conformsTo the Workflow RO-Crate (which we would have to then make Workflow RO-Crate indirection also be a workflow language!).

Note a side issue: at the moment that 5s-crate is in that case purposefully ignoring the MUST of conformsTo requirements -- we can argue over that separately!

stain commented 1 month ago

@simleo Good to go..?

simleo commented 1 month ago

@simleo Good to go..?

AFAICT, yes