Rosen, Brian wrote:
I understand why you claim such a system is a B2BUA.I can easily come up with cases where "one side fails, the other side should be brought down" is the wrong policy. Yes, it might be a good thing for some BCP.
Of course, with that example, there are no rules whatsoever
on a B2BUA, except that each side is a UA.
Does that imply that there should not be any mention
of B2BUAs at all in any document other than 3261, which
is quite deliberately vague?
I would AT LEAST want to point out issues, such as the "one side fails, the other side should be brought down".
By your reckoning, would that be out of scope for a standards
track document, and merely notes in some BCP?
I've been seeing lots of B2BUA "misuse" lately. The commonMaybe we really do need to define the fabled "transparent B2BUA" to cover this case. But I suspect if we do that we will find disagreement on exactly how transparent it should be.
one is "I have a proxy that modifies headers, therefore
I call it a B2BUA, and everyone is fine".
Better to let people hang themselves with the rope we have given them. We can fix if we find too many dead people.no protocol police, and there are no "architecture" police, but I think it is in the community's interest to at least provide expert guidance on how B2BUAs 'should' work. This is an example. 'Should' is okay with me; it would let your example be "legal".