Tom White

Any thoughts from @dask/array maintainers on the best course of action here?

> perhaps considering making Dask arguments such as `chunks=` [keyword-only](https://peps.python.org/pep-3102/) may be worth it That would be a good goal to have I think. > A while back I designed...

> I'm reading about how reduce jobs work in lithops but I don't fully get it. Is it the case that we need each serverless worker to write out the...

This has a lot of overlap with #520

It would be good to restart this and get it published. It currently write references to Kerchunk format, but is this what you would recommend @TomNicholas, or is Icechunk a...

Opened #675 and #676. See also https://github.com/cubed-dev/cubed/blob/main/api_status.md for current status.

> One approach we might take is to link the repository to [Zenodo](https://docs.github.com/en/repositories/archiving-a-github-repository/referencing-and-citing-content) and then add a [citation.cff](https://citation-file-format.github.io) file? It's slightly annoying that Zenodo issues a new DOI for every...

This is failing for Windows on Lithops, since Lithops doesn't support measuring memory on Windows (see https://github.com/lithops-cloud/lithops/blob/fdb6432e9fdc1191a4205579efc0bac32bd8daf7/lithops/worker/utils.py#L151-L152). In addition, it seems to be returning "None" (as a string), which looks...

@d-v-b in v3 chunks are not cached though, so it's much slower, which is why I opened this issue. This is the worse of both worlds as iteration works, but...

> So perhaps we bring it back but just add the chunksize checks to avoid potentially large memory usage. That would be good.