Jeremy Maitin-Shepard

> Thanks for sharing @jbms! I am currently reading and digesting that proposal. > > One point of clarification: storage transformers can act on any of the items in the...

As part of the tensorstore project we have developed a key-value store database format called OCDBT (Optionally-cooperative distributed B+tree) that can sit as an adapter between an array format like...

I think this is definitely outside the domain of the zarr v2 spec since it does not specify anything about the underlying storage, and I don't think we're inclined to...

I think the geo-zarr community is working on standardizing some http-related things for zarr v3. It could indeed make sense to say something about this issue in a zarr http...

I don't actually have much experience with zarr groups --- I've always just used lone arrays which may be organized in a directory hierarchy, and neither tensorstore nor neuroglancer handles...

Note that this is a feature supported by the neuroglancer precomputed format and has proved to be quite useful there. OME-Zarr seems to be focused specifically on microscopy imagery-related use...

@jakirkham You raise a good point that we should consider more generally what types of coordinate transformations make sense to support. I hope we can agree on the following criteria:...

@d-v-b It is true that changing the semantics of `__getitem__` would make it deviate from the norm set by Python sequences and numpy arrays (which was then copied by some...

I have not actually ever used NetCDF but I have read about it a fair bit. My understanding though is that it does not actually support non-zero origins as is...

> WRT netcdf, this does affect coordinate variables since I think that the coordinate variable has to be checked to ensure that it has the same offsets as the variable...