IETF#53 meeting in Minneapolis: 17.03.2002 - 22.03.2002 NEXT STEP IN SIGNALLING WG 1. Draft-ietf-nsis-req-00.txt: Requirements for QoS Signaling Protocols Markus Brunner Marcus Brunner went through the requirements based on the discussions in the mailing and the outcome of a 'brainstorming' with authors of some of the drafts. The biggest problem of the draft currently is that scope is very wide. Also we should get rid of application level QoS and low level QoS in order to have more focused target for the NSIS work. Risks in the current situation are that discussion is unproductive and there is a risk of over specifying. In general, it was proposed to select use cases and then create list of requirements and eventually relates them to "parts of the network" part. Also it was proposed that scenarios section and structure of requirements section remain in addition to adding description of "parts of the network" to section 6. Change the tables with MUST's, SHOULD's and MAY's to relate to parts of the network instead of relate them to scenarios. Parts of the network will not be 1) host to first hop (also first hop to host), 2) access network or 3) core to core. It was commented that it is important to differentiate the directions. It was criticized that dividing issue into sections might not be good idea. However there will be different requirements in different contents. It was pointed out in the presentation that some use cases like Voice over IP scenario (from the draft-partain-nsis-requirements-00.txt), the case when host request QoS from the network without mobility, wireless etc. and using signaling to set up the pipes. Regarding Mobile IP, the hand of decision and trigger sources should be out of the scope of the NSIS. Thus the WG should rather focus only on "establishing" QoS after hand off. Marcus presented also some concerns and pros related to Framework and it was proposed remove the figures, since they are seen as bit misleading. The changes to the text are needed in such a way that we do not want to impose any architecture or constrains. Also the intention is to try to phrase the requirements without relaying too much on the framework. In the issue, who will initiate signaling, accounting was one key issue mentioned, thus it was commented that we should not link signaling protocol solution to accounting. Also it was proposed that we don't assume that requirements are being considered to be for bi-directional reservations. The biggest discussion arose on the globally defined services and especially creating NSIS QoS classes. The question was who should define those QoS classes, since NSIS was not seen as an appropriate group to define them. NSIS could rather collect them, although something need to be done in NSIS WG as well. Massive non-interoperability problems were foreseen as well. Actually this part of the presentation raised many concerns among the DiffServ, RSVP and Intserv pioneers and it was commented that there is not technical problem at all, thus this discussion is purely political. At this point it was mentioned again that the scope and charter is definitely too wide. Also it was clarified that intention of the WG is not to define any new QoS reservation mechanism, thus we are focusing to signaling mechanism only in NSIS. Previously IP fragmentation was proposed to be one requirement, but eventually it was proposed to remove it since it may increase complexity. Some text needs to be added to avoiding duplication, ability to signal lifetime of reservation and independence of reservation identifier sections. The security requirement for the NSIS was also discussed. It was commented that more analysis about the threads are needed and current draft needs more text for security section. The next step of this work is that next version of the draft will be submitted by the end of the may. 2. Draft-fodor-reqts-cellular-netwks-00.txt: NSIS QoS Signaling Requirements from a Multi-access Wireless Perspective Gabor Fudor The presenter emphasized the importance of multi-access wireless networks. The requirements presented in the drafts are 1) host to edge signaling must be per flow, 2) QoS information, 3) bi-directional and asymmetric reservation, 4) local reservation must be supported, 5) multicast, 6) local control, Dos, 7) support for adaptivity. No comments were provided in the meeting. 3. Draft-buchli-nsis-req-00.txt: QoS signaling requirements, interfaces and architecture Sven Van Den Bosch Main goal of the draft is to enable hard guaranteed per-flow end-to-end QoS. Relation to NSIS WG draft was elaborated in the presentation. Basically the terminology and requirement section are inherently covered currently. QoS signaling and QC must not constrained to be in the data path and 2 level QoS granularities are emphasized. Mobility, modularity, universality of signaled parameters, communication between QI/QC and local administration are de-emphasized or not even applicable. Some requirements are new: priority, grouping of micro-flows. The authors of the draft support proposal to split work into three areas and propose to add use cases, terminology and requirements to WG draft. Also it was proposed to move some use cases to framework document. Some kind of document that is describing the use of NSIS would be good to have. 4. Draft-hancock-nsis-framework-00.txt: NSIS Framework Issues Robert Handcock It was stated especially in the presentation that AAA/policy capability is needed. Anyway the level of Security needs more discussion. Presentation covered issues like receiver-initiated reservations, bi-directional reservations and end-to-end signaling. Correlation between application level signaling was mentioned in the discussions. Eventually the conclusion was that the authors do not see any showstoppers, although AAA needs to be investigated more. 5. Next Steps all · Re-publish requirement document · Start working on analyses of existing signaling protocols document · Start work on framework signaling document · Possible interim meeting: target date would be in May, Location open (possibly in Helsinki, Dallas, Boston, ...)