SRP BoF - 46th IETF, Washington D.C. 11/10/99
Chair: David Tsiang, cisco Systems
Scribe: Charles Mujie, cisco Systems
Presenters:
- David Tsiang
- Peter Lothberg, Sprint
- George Suwala, cisco Systems
About 100 people attended the BoF. Amoung the attendees are folks from Cisco, Sprint, Ebone, Juniper, Verio, Lucent (ex-Ascend/ex-Nexabit), Nortel, Ixia, Cypress, Fujitsu and Pentacom.
Agenda
Brief overview of SRP (D. Tsiang)
- Brief introduction to SRP.
- Had a brief discussion on the various features supported by SRP.
Deployment & implementation (D. Tsiang)
- Mentioned some customer who have tested and deployed SRP.
- Also mentioned that we are working with some system and ASIC vendors.
User perspective (P. Lothberg)
- Talked about the problem space that SRP solves in Sprints network.
- Testing done by Sprint - PoP interconnect, MAN and Exchange-point;
Multicast
- DPT deployment in Sprint.
IESG thinking of SRP (T. Narten)
- Thomas gave some background information has to how SRP got to IESG.
- The initially consensus reached was SRP is a layer 2 protocol and that work should be done by IEEE as oppose to the IETF.
- Having had more discussion on the matter there appears to be areas of SRP that could be standardize by the IETF.
IEEE work (D. Tsiang)
- L1 SONET SPE (e.g. STS-12c), other transport
- L2 Packet delineation, CRC, frame format and SRP-fa
Peter commented that L1 work mentioned by D. Tsiang has been done by ANSI and that there is nothing else to do. On L2 packet delineation, scrambling, CRC has also been done. What needs to be standardized is the SRP frame format and fa.
IETF work (G. Suwala)
- IPS Wraps, triggers, event hierarchy, signalling, L3 interaction upon failure detection and ring concentration.
- SRP node identification
- SRP specific MIBs
- IP packet basics encapsulation, priority mapping, ARP, ring selection (choice between L2 or L3) and topology discovery.
- Modification to IGP over SRP ISIS and OSPF
- MPLS over SRP
- MPLS TE over SRP
Thomas Narten commented that any modification or work required on IGP, the place to bring that to is the routing protocol WG. They are the best people to deal with that matter.
IETF/IEEE relationship (D. Tsiang)
- Official liaison
- Dual representation at both IEEE and IETF meetings
Some of the audience felt that it is premature to establish a WG before IEEE has done the standardization work. Some felt that it should be a joint effort between the IEEE and IETF. Case in point: If ION was establish and earlier ATM won't be that screwed up."
Time frame of work and output (D. Tsiang)
- Produce Internet drafts in parallel with IEEE work
- Publish informational RFC in parallel with IEEE work
- Publish proposed standard RFC that refers to IEEE work/standards.
Thomas Narten commented that the IETF needs to work closely with IEEE on the standardization process. This apparently has worked well in the past.
Comments and Question from the audience.
1. IBM worked on FDDI II and either has or applied for a patent on Spatial Reuse. Did we know about this? D. Tsiang's responds was we didn't and thatwe'll look into it.
2. Confused between DPT and SRP. What is DPT and what is SRP? D. Tsiang's responds was DPT is the marketing name for SRP. SRP is the protocol.
3. This question came from a Nortel guy. As it is Nortel specific we didn't quite have an answer for him. What is special about SRP to doing ethernet on rings or SONET rings?
4. A user felt that the same can be achieved with POS and MPLS and MPLS TE. He wants to know why we need SRP. G. Suwala's responds was it provides an alternative to users and that there are different ways to solve a problem.
5. As SRP is a layer 2 technology most of the work should be done in IEEE 802. They have more expertise in that area than the IETF.
6. As IPS and SRP node identification is coupled with SRP, standardization work should be done by IEEE as well.
7. Question on IPS. Is the IPS protocol tied to SRP or is it stand alone and can it be applied to other technologies. G. Suwala's responds; today it is tied to SRP but there is no reason why it can't be applied to other technologies.
8. On the issue of patents what is Cisco's position. D. Tsiang's responds; similar to MPLS and MPLS patent.
9. What is the significant of the RI-bit in the SRP header? G. Suwala's responds; used to determine which ring was the packet originally sourced on.
10. Is there and SRP mailing list? D. Tsiang's responds; No. We'll take that as an action item.
11. Request for a detailed documentation that would compare and contrast MPLS to SRP prior to setting up a WG and also disclose and clarify the legal stuff i.e. patents.
12. Need to have a detailed discussion on the mailing list before a WG can be setup. Thomas Narten also added that a charter is required to setup a WG.
13. A suggestion was to setup a WG to study running IP over smart rings. I.e. the WG should decouple itself from SRP.