Excerpts from internet.ietf-822: 20-Sep-91 Re: New-ish idea on non-asc..
"Alain FONTAINE @sri.ucl (2401)
Hey, what software did you use to prepare this header ??? It obviously
'forgot'
to quote the local-part and to put backslashes where needed !!!!!
Correct mistakes. Throw again at parser. Result :
Input string : "Nathaniel S! > ? - ( : , _ \N / ; < Borenstein"
<"S!>?-(:,_\"/;<"@my.host.name> (S! > ? - \( : , _ " / ; <)|.
Aside from the trailing "|." which seems to have appeared out of
nowhere, this is indeed legal. (And, I'm happy to report, Andrew parsed
it correctly -- kudos to Everhart & Rosenberg.) However, the real
question is indeed about the software that prepares the header. In this
scheme, that software needs to start with a knowledge of the non-ASCII
character set, convert it to mnemoic, and THEN quote it past the 822
quoting rules. This is *known* to be an area where people screw up all
the time. Call me paranoid, but I've probably sent out DOZENS of
letters, over the past five years, to postmasters at sites where some
broken UA was generating From fields that looked like this:
From: John Q. Public <jqp(_at_)wherever(_dot_)com>
instead of
From: "John Q. Public" <jqp(_at_)wherever(_dot_)com>
If implementors can't be depended on to quote phrases like this, I don't
want to count on them to properly quote mnemonic non-ASCII text in
places like this either! Nor do I have much faith that there isn't a
lot of software out there that will break on "pathological" cases like
the one Alain & I managed to parse.
I agree that separating this information into a separate header causes
serious problems in matching the two forms of the address back together
again. That is indeed troublesome. I suspect that it could be solved
by having a token that appears in each & can be matched, if we really
want to do so.
I still think the bottom line is we don't have ANY great solutions to
this problem. I still like Real-XXX better than the alternatives. --
Nathaniel