Keith Moore <moore(_at_)cs(_dot_)utk(_dot_)edu> writes:
it would be interesting to know what the problems are, especially since
IDNA or something close to it seems likely to be used in email
addresses.
I'm not at all sure that I agree with Charles's blithe assertions about
the opinion of USEFOR, but the one objection that was raised to IDNA is
that it has some potentially somewhat unfortunate side effects when it
comes to Usenet feed patterns. Right now, it's common to attempt to
exclude certain types of binaries with patterns like @*mp3* and @*dvd*; so
long as all newsgroup names are ASCII words, this works fairly well, but
once you start using an encoding like IDNA, the chances of odd false
positives and negatives is quite high.
This problem may apply to any ASCII encoding system, however. (I believe
this is mostly an argument for using UTF-8 on the wire and only encoding
at boundaries.)
I don't recall any specific objections to using IDNA as the boundary
encoding for newsgroup names, and it seems a logical fit given the use of
newsgroup names in envelope recipients. (Although I don't know if the
Usenet . to - mapping for relayed messages to moderators would cause
problems.)
do we really want usefor to have to deal with 3 or 4 separate schemes
for encoding text that appears in message headers (2047 for text, IDNA
for email addresses, something else for newsgroup names?)
I certainly don't.
--
Russ Allbery (rra(_at_)stanford(_dot_)edu)
<http://www.eyrie.org/~eagle/>