On Tue, Jan 19, 2016 at 4:52 AM, John Cowan <xxxxxx@mercury.ccil.org> wrote:

[...] hash-salt should be a macro so that it can be implemented efficiently on
Schemes that are dominated by performance, as Alex Shinn noted.

Well, I had suggested this for bounds.  For salts you indicate the
value may change, so while it could still be syntax, it could not
"provide the current value of the salt" as you wrote, it would have
to expand to a level of indirection.  Even if fixed per run of the
program, you would run into issues with separate compilation.

I would leave off any discussion of using (global) salt to generate
a family of hash functions.  You provide no user API to do so, and
the practicality of implementation support for this is unclear.  It would
have to be a parameter, which would be relatively expensive
compared to eq?/eqv? hashes.