Michael Sumner

> I actually use that issue as a feature when doing things like > > SELECT ST_something() AS geometry, * FROM ... > > so that the columns with the...

Great name! I call domain "grid" in tidync. I should rename that. "Shape" is not a construct in netcdf but some docs do refer to it (a set of specific...

it did seem a bit inconsistent across the backends, I appreciate it's all changed quite quickly and settling around the obstore abstraction certainly seems good. I don't care about my...

Equivalent to 1) above: a) compact representation of the xy coordinates (i.e. bbox, or affine transform, or their "zero" equivalents 0,ncol,0,nrow) b) crs of that xy space (if describe-able, as...

I've had good chats with GeoZarr folks and earthmover folks, specifically @briannapagan and @rabernat My take, and I think they agree, is that xarray PR 9543 is the crucial one,...

I guess we could reuse the old DODS driver for pulling in virtualization manifests? Just off-cuff idea as I think of it. The real issue is this community being bound...

I was wrong, GDAL does work with the virtualization: ``` "ZARR:\"/vsicurl/https://projects.pawsey.org.au/vzarr/oisst-avhrr-v02r01.parquet\"" ``` I'll try to unpick the mistakes I stated above, I'm working on examples.

I've been told that GDAL should establish the one best way to encode transform and crs in Zarr ... and similar machinations played out with geotiff in previous years. Not...

Just discovered this is an actual driver in development: https://github.com/EOPF-Sample-Service/GDAL-ZARR-EOPF Worth exploring 🙏

I'll PR in the crs and gt handling, looks like this will be a defacto standard. Also I'll push harder at Zarr spec itself to capture this