On Thu, Jul 14, 2022 at 12:07 PM Marc Nieper-Wißkirchen <xxxxxx@gmail.com> wrote:

In these certain circumstances, it should be allowed for an implementation not support comparator-hash-function and comparator-hash.

If you don't want a hash function, specify it as #f. If you request a hash function, the implementation is required to give you one, even if it's not the one you asked for.

What does it actually mean in the context of SRFI 128 to ignore the user-specified hash function?

For example, the eq-comparator could use addresses as hashes, though it must be careful about trusting them.

I think that in order to make such hashes generally useful, an implementation will need to provide finalization queues a la Java.