Hi – thoughts on RFC 9 (zipped Zarr)?

[lubianat]

There is an open RFC on Zip support for OME-Zarr (https://ngff.openmicroscopy.org/rfc/9/)

as you are already dealing with zips and OME-Zarr, you may be interested in that RFC.

(ping the RFC author @jwindhager and @clbarnes, who also has delved into the RFC-9 discussions)

[assadiab]

Thanks @lubianat !

This is exactly the use case my viewer's ZipStore is built around, so RFC-9 is very relevant to me.

Two things already line up with the recommendations:

• STORE mode (no ZIP compression):
  I zip with no compression for the samereason the RFC gives : Zarr chunks are already codec-compressed, so ZIP deflate would just double-compress and force a wasted inflate on every read.

• Lazy per-chunk reads:
  I index the central directory once, then each chunk is a stateless Blob.slice() of the entry's data — no reader lock, no CRC pass, no decode pipeline on the STORE path: https://github.com/assadiab/Viewer_interactif/blob/c619a0fe0e40205fef9596dbb9c3677a9ae8a870/vole-core/src/loaders/zarr_utils/ZipStore.ts#L121-L142

Where my current output doesn't match the spec yet, which might be a useful real-world data point:

• My data is still OME-Zarr 0.4 (Zarr v2), so the root metadata is .zgroup/.zattrs rather than a zarr.json.
  That's because the zarr is produced by ilastik, which currently exports 0.4 only — I re-package its output downstream.
  My reader keys on whichever metadata is present, so it's version-agnostic, but I'm planning to migrate my packaging step to v3 (and then .ozx).

• For the same reason the Zarr root currently sits one folder deep in the archive rather than at the ZIP root, so my reader auto-detects the root prefix instead of assuming it's flat: https://github.com/assadiab/Viewer_interactif/blob/c619a0fe0e40205fef9596dbb9c3677a9ae8a870/vole-core/src/loaders/zarr_utils/ZipStore.ts#L156-L173

Since I control the packaging step, aligning it with the spec (flat root, v3zarr.json, .ozx, archive comment) is on my list — the reader could then drop the root-detection scan entirely.

Happy to write up what I've run into supporting both layouts on the RFC thread if that's useful.

[lubianat]

@assadiab yes, these thoughts are definitely useful.

It would be great if you could share the same thoughts either at https://forum.image.sc/t/rfc-9-zipped-ome-zarr/117630 or, if you are feeling like it, you could make a formal comment in the RFC:

e.g. https://ngff.openmicroscopy.org/rfc/9/comments/1/index.html

You can make a PR at https://github.com/ome/ngff with the comment, for example. It is open to anyone. It is a good way to make sure your thoughts (and your work) are noticed and taken into consideration :)