Hi.
This discussion is very interesting. It is clear that
considerable experience, as well as considerable opportunity for
theorizing about edge cases, has accumulated since RFC 3974 was
completed. It would be a pity to lose whatever wisdom that
experience to the back-and-forth of the mailing list.
Could I suggest that someone, ideally someone with some of that
experience first-hand, take the lead and start drawing an I-D
together that builds on 3974, discusses the configuration issues
(and traps) as they are understood today, and moves toward a BCP?
If Tony, Lisa, and the group were amenable, I'd be delighted to
slip a sentence into 2821bis that indicated that further work
was going on in this area with a "work in progress" pointer to
an I-D. Of course, first there would have to be the clear
intent to do work, not just exchange comments on the mailing
list, and an I-D.
john