I see it as the same issue because "string value" doesn't have (much of) a place in the nomenllature of Perl. I mean in typical usage; in the docs; in tutorials an well-known books. People are educated to think of $var as a scalar (which is a scalar etc.), and then when they use a reference as a hash key it loses its magic and they are suprised.

There are other places where stringification occurs. For example, if you create a file from a variable that holds a reference, and then use readdir to attempt to recover it, obviously it's going to lose its magic. I don't think that suprises anyone as much <g> but that's precisely because you are unPerlifying your data.

I take back my claim about the genesis of string-only hash keys. You're right: I don't know how the decision was made either. We both recognize simplicity; my gripe was about consistency (which, when lacking, often causes loss of simplicity elsewhere).

(Paranthetically: in Java, references can be hash keys. Of course, Java hashes are nowhere near as elegant to use as in Perl, so I don't know whose point of view this supports. Then again I think we've more or less come to see eye to eye on this, no?)