[Css-csts] procedure state transition following local detection of protocol error

[John Pietras]

Members of the CSTSWG ---
In the Crystal City meeting we discussed the inconsistent resultant
state when a procedure detects a protocol error and raises a peer abort.
The specific example cited was when a START is invoked for a procedure
that is already active. Of the Stateful procedures:

- The Buffered Data Delivery procedure raises the peer abort, but shows
no transition, which implies that it stays active;

- The Unbuffered Data Delivery procedure raises the peer abort, then
transitions to the inactive state;

- The Cyclic Report procedure inherits the Unbuffered Data Delivery
procedure's state behavior;

- The Throw Event procedure's behavior is as yet undefined (it doesn't
appear that a Throw Event can raise a peer abort);

- The Data Processing procedure raises the peer abort, then transitions
to the inactive state;

- The Notification procedure raises the peer abort, then transitions to
the inactive state.

Of course, as soon as the Association Control procedure acts on the
raised peer abort, the procedure will terminate anyway, so whether the
procedure momentarily becomes inactive may be academic. But it is my
recollection from the meeting that we agreed that for consistency all
procedures will transition to the inactive state when they detect a
situation that causes them to raise a peer abort. Am I remembering
correctly? 

Best regards,
John