This issue came up months ago in the alias, but I think it got
lost from Johns otherwise reasonable outline of the "headers
in mail messages" alternatives.
If we allow people to localize their names, then there is
every reason to believe that two people corresponding might
use different character sets to represent their names.
For example, A french user might use ISO-8859-1 to represent his
name, but might be sending a japanese message (with a japanese
subject line) encoded using ISO-2022. Unless we have a univeral
character set (like mnenonic, iso10646, or unicode) then we need
the ability to represent character sets and encodings on a
per-header line basis.
This leads me to like Nathaniels Real-xxxx proposal.
Neil