>> Thinking about it some more, I am wondering: is it more important
>> - to distinguish keys which only differ by their catcode, or
>> - to allow for any token as a key (e.g. Hef{}feron, which currently
>> breaks the delimited argument approach) ?
> Hello Bruno,
> I'd say that at the very least we should _store_ tokens and not
> _strings_. So category codes should be preserved when putting stuff in
> or getting them out. After all, sequences might be used for all sorts of
> things, and the tokenization may be important.

Of course, the <value> must be a token list, but the <key> could be
more restricted?

I'm guessing that the use of a prop may be

\q_prop name \q_prop {S\o m\c{e}$t_{hi}n^g$}
\q_prop country \q_prop {Br\'azil}
\q_prop I.D. number \q_prop {2CUOHE@#@}
\q_prop ef{}ficiency \q_prop {12}

Here <values> may be anything, but <keys> are rather well behaved. The
key (sic) property of <keys> in my understanding is whether they
differ or not, and detokenizing will not cause too many collisions?