Currently the compression algorithm for caches is chosen implicitly according to what tools and versions are available on an environment.
It would be nice if it were possible to instead manually select the compression algorithm, including choosing no compression algortithm.

The use-case is, that I build something on one environment and want to test it on multiple others.
I don't want to use artifacts for that, as that is not really their intention imho and you also cannot delete the artifacts.
Imho the cache with the workflow run id as cache key should be better suited for inter-job file sharing.
But you currently have the problem, that if you for example build on a gzip environment like windows-latest and then try to restore on a zstd environment like windows-2016, you get a cache miss.

#543 would improve on that example, but as soon as you swap it around and build on windows-2016, wanting to test on windows-latest, #543 would not help.
So it would be nice if you could manually select the compression algorithm in the cache-action (and therefore having it possible here first), to be able to choose the lowest common denominator (including no compression) that can be understood on all part-taking environments.