RE: Notices to ASPTM message mapping
"Kolomaznik Jan" <[email protected]>
| Newsgroups | gmane.ietf.sigtran |
|---|---|
| Message-ID | <[email protected]> |
Ad 1) I think, it will be better if any recommendation word (like SHOULD) will be used in talking about a possibility of ASPTM messages mapping on any SCTP stream carrying data traffic. I agree that in situations where more than one data traffic streams are used the minimization of possible packet loss cannot be so high, but in cases where only one data traffic stream is used it is 100%. Ad 2) If ASTPM Ack messages will use a data traffic SCTP stream, I do not see any practical reason for this type of restriction. - But if ASPTM Ack messages will be transmitted on SCTP stream 0, there is a theoretical possibility that ASPTM message will be evaluated earlier than previously sent data message(s). This possibility could happen both in ASP Active and ASP Inactive procedures. So my meaning is, when a ASPTM not Ack message used data traffic SCTP stream, the ACK message should use data traffic SCTP stream also. The second thing is that from protocol theory point of view the Ack on different stream than acknowledged message looks not good. So if there is not any practical reason for changing of stream for Ack message, the Ack message should use the same stream. Jan Kolomaznik -----Original Message----- From: Brian F. G. Bidulock [mailto:[email protected]] Sent: Monday, May 29, 2006 8:21 PM To: Kolomaznik Jan Cc: [email protected] Subject: Re: [Sigtran] Notices to ASPTM message mapping Kolomaznik, Kolomaznik Jan wrote: (Mon, 29 May 2006 13:16:15) > > Hello, > > I would like to share any findings we have made during implementation > of SUA in IPSP environment. Our work has been based on RFC 3868 and > draft-ietf-sigtran-sua-implementor-guide-03.txt. Please send comments > if you find out we are not true. > > > 1) Problem with mapping of ASPTM messages on SCTP stream. - Which > stream number to use? ... there are two relevant paragraphs in RFC > 3868, sua-implementor-guide-03 does not solve this question. > > RFC 3868, section 4.2.1: > > "All non-Transfer messages, with the exception of ASPTM, BEAT and BEAT > Ack messages SHOULD be sent on SCTP stream '0'. ASPTM messages MAY be > sent on one of the streams used to carry data traffic related to the > Routing Context(s), to minimize possible message loss." > > RFC 3868, section 4.3.4.3: > > "It is possible for the ASP to receive Data message(s) before the ASP > Active Ack message as the ASP Active Ack and Data messages from an SG > or IPSP may be sent on different SCTP streams. Message loss is > possible, as the ASP does not consider itself in the ASP-ACTIVE state > until reception of the ASP Active Ack message." > > > So ASPTM messages MAY be sent on any SCTP stream, but generally it is > better to send they on any traffic stream (not on SCTP stream 0). It > practically means, a sending entity MAY send it on any SCTP stream and > a receiving entity SHOULD be prepared to receive it on any SCTP > stream. Yes. In general, an implementation should be prepared to receive any message on any stream. > > > 2) Mapping of ASPTM Ack messages on SCTP stream ... I think it is > general principle to be received message acknowledged on the same > stream as an acknowledged message has been received - but this general > principle is not enforced in RFC 3868 or > sigtran-sua-implementor-guide-03. This situation may leads to the > situation following the next example: > > ------------ ASP Active/SCTP stream 1 -----> > > <-------- ASP Active Ack/SCTP stream 0 ---- Yes, that is possible. Did you see any reason why it would be necessary to restrict the situtation further? --brian -- Brian F. G. Bidulock [email protected] http://www.openss7.org/