Automated Review URLs

• Readthedocs
• render latest/index.bs
• diff latest modified

(Note for maintainer: please squash when merging)

Thanks for your comments @btbest!

I'm not 100% sure why case-insensitivity is important though? I would rather postpone that decision to a future spec change, if needed.

Definitely on board with unique names though...!

I'm not 100% sure why case-insensitivity is important though

I consider axis naming to be human, and humans don't think of "frequency", "Frequency" and "FREQUENCY" as different names.

On second thought, I think requiring lower case is probably bad: cost is potentially high (vendors might not like having to map their custom axes to a lower-case version for OME-Zarr) and benefit low (prettier metadata I guess? marginally simpler for implementations).

I still think requiring case-insensitive uniqueness is probably good: cost is zero (just add a suffix; you're using a non-standard axis name anyway) and benefit is moderate (developers won't have to decide whether "x" and "X" are different axes or the same; users should see more interoperability between tools). In particular, for the single-letter case, I consider it a benefit that this would forbid using lower- vs upper-case naming according to an existing convention for expressing transformations. If it turns out sensible to lift this restriction for RFC-5, it would be less painful to keep the restriction now and ease it when it makes sense, rather than the other way round.

Requiring case-sensitive uniqueness would also be fine. At least we'd have a standard. Developers can know what to expect and choose to be strict or permissive at their own risk.

I still think requiring case-insensitive uniqueness is probably good: cost is zero (just add a suffix; you're using a non-standard axis name anyway) and benefit is moderate (developers won't have to decide whether "x" and "X" are different axes or the same; users should see more interoperability between tools).

I think the spirit of RFC3 is to not tell people what to name their axes. And if developers want to aggregate across datasets that use mixed-case names like "X" and "x"`, those developers can make the names lowercase before doing the aggregation. So I second the idea to push this to a later spec change.