[PROPOSAL] re: closing cords

[Dave Kormann <davek@research.att.com>]

well, not REALLY a proposal, but:
 > i would prefer that we not require close messages both ways.  in this
 > case, we should both (a) state that explicitly, and (b) include a note
 > to the effect that cord types that require more complex shutdown
 > behavior should define it for themselves.
 > 
 > note that even assuming we don't _require_ close acks, you still
 > _might_ get a close message on a cord you've already closed, due to a
 > race condition.

i added this stuff in the form of the paragraph:

Note that cord-closed messages need only be sent in one direction, by
the endpoint closing the cord, and once the message has been sent, the
cord should be assumed to be closed. Implementations should not expect
to receive a cord-closed message in reply. However, due to potential
race conditions, such messages may be sent both directions. As with
any message on a closed cord, these messages should be ignored.

feel free to complain about the wording.  i'll probably add
"extraneous" to the phrase "these messages should be ignored".
				dk