Open effigies opened 1 day ago
I've created maint/1.10.0: https://github.com/bids-standard/bids-specification/compare/v1.10.0...maint/1.10.0
I believe this contains the semantically neutral differences that have been merged since the release of v1.10.0. Please let me know if you disagree. You can also see the changes against master
, which I didn't include: https://github.com/bids-standard/bids-specification/compare/maint/1.10.0...master
I will start working on the tooling to handle the schema publication from the maintenance branch.
The BIDS specification, schema and validator lifecycles are currently mixed. The schema is updated linearly with main, and there is no method to maintain a branch of the schema that fixes issues with the schema for the stable release without also including new features in the
main
branch.As the validator is now heavily reliant on the schema, there is no difference to the validator between an update that fixes a bug in the schema and one that introduces a feature, which could potentially be changed or reverted before the next release of BIDS.
Here I propose a parallel branching strategy for the specification and the validator.
Legend:
bids-specification:main
bids-specification:maint/1.10.0
bids-validator:stable
(orbids-validator:maint/1.15.x
)bids-validator:main
By using
.postN
releases for fixes to the schema and.devN
releases for new changes, the stable validator could follow the latest.post
and the development version could follow the latest.dev
.I also include a
v1.10.0.post1
tag, which could be used for schema fixes that have rendering consequences, or things like fixing typos in the spec.Note that the "merges" between the specification and validator branches are not actual git merges. They would be the validator updating pins to versions of the schema pushed to https://jsr.io/@bids/schema/versions.
This post is motivated by validation considerations, but has implications for how we operate on the specification. Therefore I am posting this here, as well as https://github.com/bids-standard/bids-validator/issues/85, where I expect the discussion to focus more around the mechanics of keeping validator branches synchronized to schema releases.