In the past the IETF tried to discuss in a green-field manner how to internationalize email addresses and other email headers. None of the many options seemed really low-risk, however, and those discussions didn't get very far -- rat-holes being very frequent, and people having very different notions about what a solution could look like and what the outcomes and side effects might be.

The BOF I attended last week, however, was much more focused: it was about whether or not the IETF should charter a group to work on a specific experimental solution for internationalizing email addresses as outlined in several draft documents. That solution focuses on SMTP, so that "consenting adults" in an environment that wants to do i18n addresses may do so, and SMTP agents have a reasonable way to deal with these addresses when communicating to the wider world. It's a somewhat transitional approach, acknowledging that existing Mail User Agents (MUAs) and Mail Transport Agents (MTAs) won't immediately be able to handle these.

The discussion was mostly abour risks and unknowns. Some of the risks:

• It's not known how (and when, and by whom) IMAP would be updated to handle i18n addresses, and how clean that could be.

• It's not known how POP would be updated to handle i18n addresses (Chris Newman stepped into the line of fire here)

• It could be difficult to manage VCards, iCalendar objects, and Web pages where mailto URLs and email addresses appear. Some of these support i18n, some don't; even ones that do may have incompatible representations.
  

• Although i18n addresses are supposed to remain within these groups of consenting adults and be transformed before transmission to non-i18n MTAs, it's not known to what extent these would actually leak out.

• When these do leak out, the user experience of email users with non-i18n MUAs could be unsatisfactory.

• There will probably be serious difficulties when non-i18n MUAs are used to try address mail with i18n addresses -- it's possible that sometimes only an i18n address is known and the user can't figure out how to enter it or isn't allowed to by their software.

Normally the IETF is rather risk averse when there are so many unknowns, particularly when 'Net fragmentation might occur. In the past the IETF has gone around and around trying to get more certainty before even chartering a working group. This time, however, since there have been so many discussions and stalled related efforts before, the attendees took a leap into the unknown and approved the working group -- unanimously, I believe. Imagine Admiral Farragut's sailors taking a "hum" and deciding together to damn the torpedoes.