Feature Request: use `ngff-zarr` for the V3 writer metadata structures.

[BrianWhitneyAI]

We introduced Axes and Channel classes to help with metadata organization. We could use ngff-zarr here to improve our validation and outsource some of the formatting responsibilities.

[SeanDuHare]

Lets try to do this as a follow-up to Brian's writer migration PR

[imagejan]

Before the writer migration, bioio had already switched from ngff-zarr to ome-zarr-models (for Python versions >=3.11) in bioio-devs/bioio#113 (see also comment in bioio-devs/bioio-base#25 (comment)).

Could we maybe switch completely to ome-zarr-models? This would make it much easier to deploy to conda-forge! (Currently, ngff-zarr doesn't have a conda package, and depends on itk-wasm that's also absent from conda-forge.)

[toloudis]

I think @BrianWhitneyAI will know more details but we found that we couldn't use ome-zarr-models because its support for OME NGFF v0.5 / zarrv3 was not quite as ready, plus some other dependency issues if I remember correctly? I believe we have a weird situation with no particular "blessed" ome data model+validation package right now, and several competing ones. ome/ome-zarr-py#407 and ome/ome-zarr-py#452

[BrianWhitneyAI]

Dan covered most of the key points. In short, ome-zarr-models hasn't seen much recent development and does not yet provide the support we need for the latest format. On the other hand, ngff-zarr is up to date, so this ticket proposes evaluating whether it makes sense to adopt it.

We don’t currently have a strong preference for which library to use long-term, and the community doesn’t seem to have reached a consensus either. When building the current V3 support, I intentionally took a dependency-light approach to minimize support and maintenance issues like this. As a result, some of the metadata handling ended up being organized in a less-than-ideal way.

[BrianWhitneyAI]

we have decided to move this to backlog. We are planning to combine the V2 and V3 writers together and opt for the dependency light build. We have had similar deployment issues that block usage of ngff-zarr internally so until we resolve those internal blockers we are not planning to implement additional ngff-zarr support.

[BrianWhitneyAI]

Closing this as unplanned. Will reopen if things change.