Re: Closure on canonicalization, I hope

Accept-charset could be included as a possible shorthand for
multiplexing acceptable charset parameters for ALL types that have a
'charset' parameter, but I think it is only worthwhile if the client
accepts multiple charsets and multiple types that take a charset
parameter.

I'm not sure it's worth the complexity of the model, or whether we
want a more general way of multiplexing parameters for the types that
are named by accept clauses, and I don't think it belongs in the
standard without at least a little implementation experience.