Re: [Gen-art] Gen-ART review of draft-ietf-tsvwg-sctp-strrst-12

[Michael Tuexen <[e-mail]>]

On Nov 30, 2011, at 10:10 PM, Vijay K. Gurbani wrote:

> On 11/30/2011 02:19 PM, Michael Tuexen wrote:
>>> - S5, second paragraph --- "The procedures outlined in this section are
>>> designed so that the incoming side will always reset their stream
>>> sequence number first before the outgoing side which means the re-
>>> configuration request must always originate from the outgoing side."
>> A stream is unidirectional. So it is outgoing for one side, incoming
>> for the other. The outgoing side has to request the reset, the incoming
>> side will perform it and acknowledge that the reset has been performed and
>> then the outgoing side will also do the reset.
> 
> OK, so let's see ... assume two hosts A and B with two unidrectional
> streams open between them:
> 
> 
>    --------- a' -------->
> A                         B
>    <-------- b' ---------
> 
> a' is outgoing stream for A and incoming stream for B.
> b' is outgoing stream for B and incoming stream for A.
> 
> What you are saying is that either A has to request a reset on a'
> or B has to request a reset on b'.  In the worst possible case,
> both A and B can request a reset on their respective streams
> (A on a' and B on b') without leading to deadlock or a violation
> of the protocol.  Yes?
> 
> If so, then this discussion at least makes it clear to me.  However,
> I will leave it to you on whether or not to clarify this further
> in the draft.  When I read the draft initially, I certainly had
> enough doubts that resulted in this email exchange.
OK, what about making clearer what the incoming/outgoing side is...

What about:

<t>One important thing to remember about SCTP streams is that they are
uni-directional. The endpoint for which a stream is an outgoing
stream is called the outgoing side, the endpoint for which the stream
is an incoming stream is called the incoming side.
The procedures outlined in this section are designed
so that the incoming side will always reset their stream sequence number
first before the outgoing side which means the re-configuration request
must always originate from the outgoing side. These two issues have important
ramifications upon how an SCTP endpoint might request that its
incoming streams be reset. In effect it must ask the peer to start an
outgoing reset procedure and once that request is acknowledged let the peer
actually control the reset operation.</t>

Best regards
Michael
> 
> Thanks,
> 
> - vijay
> -- 
> Vijay K. Gurbani, Bell Laboratories, Alcatel-Lucent
> 1960 Lucent Lane, Rm. 9C-533, Naperville, Illinois 60566 (USA)
> Email: vkg@{bell-labs.com,acm.org} / [e-mail]
> Web:   http://ect.bell-labs.com/who/vkg/
>