s3fs's https://github.com/fsspec/s3fs/pull/917 was in response to the IfNoneMatch feature from the summer. How would people imagine this new feature being surfaced in a filesystem abstraction?

conda provides cudatoolkit and associated packages. Does this solve the situation?

Actually yes it does....except I seem to remember that it doesn't go back that far in cuda versions. I can't seem to find it again right now.

The condos 200-employee threshold licence change is problematic for some.

As long as you stay out of the "defaults" and "anaconda" repos, you're not subject to that license. For my needs conda-forge and bioconda have everything. I'm not sure about the nvidia repo but I assume it's similar.

Actually all CUDA Toolkit libs are already available through the conda-forge channel: https://anaconda.org/conda-forge/cuda-cudart, https://anaconda.org/conda-forge/libcublas, etc.

I'd like to point out that fastparquet has been built for wasm (pydide/pyscript) for some time and works fine, producing pandas dataframes. Unfortunately, the thread/socket/async nature of fsspec means you have to get the files yourself into the "local filesystem" (meaning: the wasm sandbox). (I am the fastparquet author)

Clarification: resistive heating is very near 100% efficient, meaning that all of the energy you pull ends up in the air or water it is meant for. Gas water heating is at the very best ~90-95% efficient (e.g., top-end condensing tankless water heaters), but often as low as 60%: plenty of heat going out the exhaust. You think resistive heating is inefficient because policy has made electricity much more expensive per kwh than gas/oil.