IETF 46 - GSMP Working Group The meeting was chaired by, and minutes are provided by, Avri Doria and Kenneth Sundell. The meeting was divided into 2 sessions. The first session was devoted primary to a review of ongoing work, while the second session was devoted to discussions of proposals submitted for improving GSMP. GSMPv3 Document Status Tom Worster started the meeting by giving a status of the GSMPv3 specification. Now that most of the charter related items have been submitted, it is time to start preparing the document for last call. The plan is to fix all of the items listed in his presentation and to do working group last call on the email list before IETF47. One of the pending issues had to do with security mechanisms in GSMP. The working group accepted the proposal that IPsec would be required for IP implementations. Since GSMP can also be implemented directly over Ethernet or ATM, IPsec would not be required in those cases. Tom will produce language on security that covers IP, ATM etc. There was also a discussion of specific MPLS labels. Though the label definition have been opened up to accept MPLS labels we still need volunteer to define these labels. Application for IANA number support Avri presented a list of numbers for which IANA application would be made. The group accepted the proposal with one modification. Group consensus was against including service configuration IDs among the externally defined numbers. GSMP MIB Hans Sjostrand presented the GSMP MIB. This is as yet a bare bones MIB which need to include information to support partitioning and partition specific information. A MIB design team was formed consisting of: Hans, Paul Jardetsky and Joachim Buerkle. As the group received information that the MSF was interested in working on the MIB as well, the liaison was asked to find out the extent and availability of people from the MSF to participate in the MIB work. MPLS Reservations There were two presentations on methods for supporting MPLS reservations. Joachim Buerkle presented is proposal for the addition of usage bits for this purpose, while Tom Worster proposed the addition of a reservation command. Joachim agreed to work put aside his proposal in favor of the new command. Tom and Joachim agreed to work together on language that would be added to the GSMPv3 specification. It was noted during the discussion that the reservations needed to be per partition. Also, some issues where discussed about how to handle the reservations when the CAC was on the controller. A follow-up MSF Liaison was presented. This was a response to the questions the GSMP WG had sent after IETF45. Second Meeting: A minimal ATM switch resource model Constantin Adams presented an ATM Switch minimal resource model. The purpose was to include a minimal resource model together with a set of messages for exposing, configuring and controlling these resources into GSMPv3. How to move this work forward? The authors were in favor for granting the document as a working group document. The minimal resource model as such was accepted, but the document consists of a number of other models as well, for example switch fabric, port and connection models. The working group chairs suggested that the document should be reworked, i.e. to leave out details of other models and submit it as working group document. The group agreed. A number of comments were raised regarding limitations on e.g. buffering policy. Discussions will continue on the GSMP list. It was also decided that the co-chairs should pre-review the rewritten draft before granting it as working group document. A QoS extension to the minimal resource set Mahesan Nandikesan presented a QoS extension to the previously discussed minimal resource model. He presented a set of messages in order to guarantee QoS for a schedulable region, which is an abstraction that captures the capacity of a multiplexer. The messages are intended to be used in conjunction with the messages provided with the minimal resource model discussed in the previous presentation. Questions were raised whether these QoS extensions should be included into the minimal resource model proposal. This was not the goal of the authors. There was a feeling from the group that the proposal included extensions to an extension, especially regarding MType negotiation during the initialization phase. It was felt that a model should be regarded as a complete package, so the question was if this should be folded into the Mtype=1. There were no objections, so in practice the work was in direct approved as a working group document. An ATM switch service model for GSMP Mahesan Nandikesan presented an ATM switch service model for GSMP. The proposed services are inherited from work done in the IEEE PIN working group. Currently IEEE PIN is using an API; this document proposes the use of GSMP instead. The authors are proposing to include the traffic and QoS parameters in a special add branch message, with input and output labels equal to zero. This change would introduce a change in behavior for this add branch message since the combination would imply; no connection. This statement raised objections in the crowd, since VPI=0, VCI=0 is a valid connection in a switch. The working group agreed that this document doesn't yet fit the service model straight jacket. The document is not approved in current shape as a working group document. Discussions will happen on the GSMP list. Adding redundancy to GSMP Kenneth Sundell proposed to add a new event message and changes to the switch configuration request/response message in order to enable controller redundancy in GSMP. The event message was introduced in order to notify the controllers that other controllers for a certain partition exists. The purpose of the new changes to the configuration message was to also give the controllers networking addresses to all controllers for a certain partition. He also proposed a partition state machine for this purpose with specific demands for a switch partition MIB. The group accepted the basic idea. However, the responsibility for controllers to find each other was not regarded to be the role of GSMP. The state machine has to be described more in detail. A new version of the draft is expected in order to reflect the suggestions that raised during the meeting. The draft will, when approved, be folded into the GSMPv3 specification. GSMP Enhancements Balaji Srinivasan presented a number of proposed enhancements to the GSMP. The proposal consists of quite a number of changes with emphasis on simplification of message parsing and improving the support for multiple partitions. During the presentation he also proposed changes for to better support multi-point to point messages and also a number of different issues that were not included in the draft but posted on the mailing list prior the presentation. There was a desire of the group that the adjacency message should be excluded from the re-hash of all messages since it currently can be accommodated in one single ATM cell. There was discussion of the proposal of changing the message format to aid in forwarding. There was no consensus reached on this issue. There was also a proposal that the GSMPPv3 specification be divided into several documents. One would contain the main part of the specification, while the others would each document a message encapsulation. There were a lot of agreements in principle but not in fact. There were issues that need to be discussed on the list. A three-week list discussion time period is set. The authors of the drafts are expected to start these negotiations. Extended move branch concept for GSMP Joachim Buerkle made a brief presentation on issues with the move command. The consensus of the group as to accept the problem as work item and to continue discussion on the list.