Re: [tsvwg] Question about RFC 4960 (SCTP) - initial stream sequence number

[Yutaka Takeda <[e-mail]>]

Oh -! how did I miss that one...
Thank you Michael, my problem solved!


On Fri, Feb 8, 2019 at 3:40 PM Michael Tuexen <
[e-mail]> wrote:

> > On 9. Feb 2019, at 00:22, Yutaka Takeda <[e-mail]> wrote:
> >
> > As I was reading RFC 4960 (also its errata-08), I was not able to find
> any text which explained
> > how the initial Stream Sequence Number, or SSN, of DATA chunk for a
> particular stream identifier,
> > or SI, is determined.
> This is specified in https://tools.ietf.org/html/rfc4960#section-6.5
>
>    The Stream Sequence Number in all the streams MUST start from 0 when
>    the association is established.  Also, when the Stream Sequence
>    Number reaches the value 65535 the next Stream Sequence Number MUST
>    be set to 0.
>
> Does that answer your question?
>
> Best regards
> Michael
> >
> > It is possible that a receiver of a stream can take the first DATA chunk
> that is "due" by Cumulative TSN,
> > however, waiting for the chunk to be due (everything else before the
> chunk was all received) could cause
> > a delay for delivering the already received chunk to the upper layer.
> >
> > Here's an example on the receiver side queue:
> >
> > TSN SI SSN BE status
> > --- -- --- -- -------
> > 100  2 10  11 missing  <== Cumulative TSN
> > 101  2 11  11 received (waiting for TSN:100)
> > 102  2 12  11 received (waiting for TSN:101)
> > 103  2 13  11 missing  (would wait for TSN:102)
> > 104  1 10  11 received <== should it wait for TSN:103?
> >
> > Initial TSN is 100 in the above example. For simplicity, both streams
> (SI=1 and 2) starts
> > with SSN=10.
> >
> > My first question is if the chunk (TSN:104) has to wait for arrivals of
> TSN 100 and 103.
> > If yes, then as I mentioned, it means there's a head-of-line blocking
> between the streams.
> > (SI=1 is blocked by SI=2 in the above case)
> >
> > If no, then we could pass the chunk at TSN=104 to "stream level"
> (reassembly queue, etc), but the
> > stream level would not know whether the SSN=10 is supposed to be the
> first chunk to be delivered to the
> > upper layer (of course, to guarantee the ordered delivery).
> >
> > To me, along with "unordered" delivery is implementation, it is natural
> to pass received chunks to stream level
> > right away regardless of "ordered" or "unordered" and guarantee the
> ordering at stream level (without being
> > blocked by other streams' retransmission), but doing so requires
> "initial SSN" agreed by both ends.
> >
> > I more than likely is missing something here.  It would be greatly
> appreciated if you could give me any insight
> > into this problem.
> >
> > Yutaka
>
>