Emmanuel Mathot

I find @d-v-b's proposal compelling - using Zarr's native JSON capabilities seems like the natural approach. But I understand @rabernat's compatibility concerns. I agree with @chris-little that we shouldn't lock...

Thanks @rabernat for this insight - you've nailed the core issue that should probably be front and center in the GeoZarr spec. You're absolutely right that Zarr is fundamentally >...

In this PR #89, alongside terminology clarifications, I also put propositions to improve the abstract and intro section to describe the purpose of the GeoZarr spec better.

Any news on this issue? xarray becomes old and difficult to use latest zarr change in projects with openeo

From this [comment](https://github.com/zarr-developers/geozarr-spec/issues/83#issuecomment-3252607981) of @christophenoel in another issue > just want to add that creating a completely new standard has been proposed in the past (including by myself), but after...

Seems that this has been already discussed extensively here too: https://github.com/zarr-developers/geozarr-spec/issues/20

Can't we add the toleration in the pod creation if the corresponding CUDARequirements is in the workflow?

> * What happens if you set a pod node selector with GPU to steps that don't need it? Isn't there a way to distinguish the steps with the requirement...

Thanks for the feedback, @Kirill888 and @mraspaud. I agree with your pragmatic approach - converting CF 8.3 to GDAL-aligned representation on first access is indeed the right starting point. You...

> Or am I missing something? You are correct. Although our use case with Sentinel-1 is covered by CF conventions, the provided GCPs actually form a regular grid over the...