IETF 68, ANCP WG Meeting minutes. TUESDAY, March 20th, 2007: 13:00 - 15:00 ------------------------------------------- CHAIRS: Wojciech Dec (wdec at cisco.com) & Matthew Bocci (matthew.bocci at alcatel-lucent.co.uk) SCRIBE: Richard Pruss (ric at cisco.com) AGENDA ------ 5 mins - Administrivia, WG Status and WG Docs Update - Chairs 10 mins - ANCP Framework / Requirements - Stefaan de Cnodder (stefaan.de_cnodder@alcatel-lucent.be) http://www.ietf.org/internet-drafts/draft-ietf-ancp-framework- 01.txt 15 mins - ANCP Multicast Discussion - Chairs 15 mins - ANCP Acess Node MIBs - Stefaan de Cnodder (stefaan.de_cnodder@alcatel-lucent.be) http://www.ietf.org/internet-drafts/draft-decnodder-ancp-mib-an- 01.txt 15 mins - ANCP Security Threats and Requirements - Hassnaa Moustafa (hassnaa.moustafa@orange-ftgroup.com) http://www.ietf.org/internet-drafts/draft-ietf-ancp-security- threats-00.txt 15 mins - ANCP Protocol Draft - Derek Harkness (dharkness@juniper.net) http://www.ietf.org/internet-drafts/draft-ietf-ancp-protocol- 00.txt 10 mins - Protocol Draft: Versioning and TLV numbering - Chairs MEETING MINUTES: ---------------- Administrivia, WG Status and WG Docs Update – Chairs ---------------------------------------------------- Woj updates on the milestones, some dates have changed, not on the ietf website yet, but it is agreed by the AD's. Update on the Framework and Requirements, accepted previously as a WG I-D Need definitions of requirements for the light weight transport protocol. Need to start that on the alias. Topic was raised at the DSLF. ANCP Framework / Requirements - Stefaan de Cnodder -------------------------------------------------- http://www3.ietf.org/proceedings/07mar/slides/ancp-1.ppt Menachem Dodge (ECI) asks about the repetition of the security requirements in framework and security drafts. Woj feels that all the security pieces should go into the security document. ? feels that including all the security document into the fw document would be too much. Mark Townsley feels that requirements should not be trying to be in two drafts. Causes excessive book keeping between the two drafts. Matthew Bocci would like all the requirements in one draft. Threat analysis will result in requirements that need to be reflected into the Framework document Hasnaa Moustafa (FT) will have all the security requirements in the ANCP WG-ID security draft. Agreed. Mark Townsley will need to lock at the two documents together when they go for last call. If the IAB area does not see a framework with the security requirements in it will not pass. Mark outlines that the security parts can be optional for use but not optional for implementation of the protocol. Woj said that the DSLF discussed the same in the last meeting of the DSLF. Input on what the optional usage is welcome from operators. Discussion on node functional requirements. The framework doc is an informational document so it does not strictly need to be very formal and avoid specifying node requirements as the point of the document is to get everyone on the same page enough to do the real work of engineering a protocol. Conclusion; the node functional requirements (MUSTs, SHOULDs, etc) should be either made lower case or made descriptive. ANCP Multicast Discussion - Chairs ---------------------------------- http://www3.ietf.org/proceedings/07mar/slides/ancp-4.ppt Matthew Bocci goes through the first 2 slides that present an overview of the mailer discussion. Various points of the meaning of multicast and ANCP discussed, mostly around mechanisms. Scot Brim (Cisco) asked a question if ACL is acting on the IGMP protocol. Answer is yes; the ACL is typically taken to be acting on the multicast control protocol. The ACL can be pre-provisioned or on-demand. Mark Townsley asked a question of what direction the provisioning was happening, is it from NAS to AN or vice-versa? Answer: NAS to AN Lei (Telenor) ask if CAC is only being discussed in the multicast context or unicast CAC services as well. Toerless (Cisco) makes the point that the CAC enforcement for multicast has to be in the network. Lei makes the point that some integration is needed as CAC needs the information of what multicast is occurring Lei points out that information delivering for what CAC information to unicast severs can be done with ANCP. Francois Le Faucheur (Cisco) advocates that even if only the multicast is address care needs to be take to allow unicast services CAC Derek Harkness (Juniper) advocates for a virtualized the subscriber on the NAS Matthew Bocci this is close to the reporting use case of ANCP Stefaan (Alcatel) questions whether unicast CAC is in the scope Francois Le Faucheur addresses Stephan's question on what is Unicast CAC. Either the NAS is involved or the NAS provides information to the CAC element. CAC for unicast and multicast needs to be integrated. Chair question to the WG on slide 3: Is ANCP’s multicast use-case requirement really about making the combination of access node and nas look like a single element to external elements? Moti (ECI) suggests that a black box approach does not go far as engineering cannot be done with this blackbox. Moti Morgenstern educates that some deployments today have access node taking multicast decisions and other deployments have NAS taking multicast decisions Moti feels that both should be supported as both are in deployment. Matthew Bocci note that both use cases exist Toerless notes replication can happen in two places and the policy choices can happen in two places. Mark brings the conversation back to the architectural guideline that does not make the information in the system worse. Roberta Maglione (TI) brings a service provider perspective. Replication on access node solves one problem (optimal bandiwdth) but introduces other (loss of functionlity) when compared to NAS based replication. Time based accounting desired with access-node replication. Moti Morgenstern makes the point that is some cases the policy can be delegated to the access node. Woj would like the WG to focus into looking as to what is the multicast use case in the network. Mark asks if the use cases to be guided by the architectural approach against the use cases. There is certainly an element of the architecture that needs to be considered. High level use-cases should be documented. Also notes that ANCP is meant to be control protocol not solely provisioning. Woj concluding on proposed WG steps for moving forward: 1. Framework text to be added on two boxes (NAS and AN) connected via ANCP that functionally look like one black box 2. For multicast section framework text needs black box level multicast use-cases that follow the principle of multicast replication on the AN but no loss of functionality. CAC is part of this, but addressing the full unicast+multicast CAC system solution is not in scope of ANCP WG. The existence of the CAC use-case is to be noted and ANCP mechanism designed to fit as best as possible. 3. More detailed solution proposals to be brought in as part of the ANCP protocol specification work. Requirement for integration of multicast CAC with unicast CAC to be added to framework text. Woj asks about about objections. No objections to these steps for moving forward. Nicolas (DT) asks a question if multicast group excludes source specific multicast Derek Harkness - where you identify a multicast group both should be identified as G and S,G Matthew Bocci; break the use cases into a small set of applications for this set of use cases. Mark Townsley suggests that whatever is missing in a use case gets transports in ANCP might fix the requirement. Woj This level of detail is probably more appropriate for protocol doc, but a note about the assumption that both are to be addressed needs to be made in the protocol spec. This is an action for the editor. Woj question to Mark. There is some work in the MBONED that covers some of the authorization, accounting and how the two groups could work together? Hiroshi Ohta (NTT - co-chair of MBONED) MBONED is gathering requirements for similar space at the moment. Would like to work with ANCP WG. Mark instructs chairs from both groups send emails to the other groups on the activity that is occurring the same space. ANCP MIB - Stefan De Cnodder ----------------------------- http://www3.ietf.org/proceedings/07mar/slides/ancp-0.ppt Question on partition id. The question is on the default value for the partition id. why do we have a default? No clear answer. Woj comments we should not exclude other methods. Moti question on partition id. Why are the id's needed. Is it possible to have controllers with different roles? Woj, partitions allow partitioning of the lines to controllers. Woj noted functional partitioning came up before and was not identified as being required. Bert asking if separate SNMP agents will be serving the partitions, or will a single SNMP agent be serving all? This has a profound effect on index in the MIBs Woj Impression is that management will be in one management domain, but this does not answer the question. Question and discussion to be taken to the alias. Stefan requests for draft to be taken as WG I-D. Call to be made on the WG alias. ANCP Security Threats and Requirements - Hasnaa Moustafa --------------------------------------------------------- http://www3.ietf.org/proceedings/07mar/slides/ancp-2.ppt Question on security requirements. Repeat of previous answer; requirements will be developed in security draft, and the framework will have a reference to the security doc. ANCP Protocol Draft - Derek Harkness ------------------------------------ http://www3.ietf.org/proceedings/07mar/slides/ancp-3.ppt Chair asks how many people have read the draft. Approx 5 hands noted. Mark wants more people to read and comment on the draft. Without sufficient reviewers doc will not progress. Chairs Doc still contains a fair amount of framework text. Also, the spec does not stand on its own – still assumes existence of gsmpv3-base-spec draft which has expired. This dependency must be removed, and spec firmed up to allow implementers clear guidance. References to expired or non-existent drafts removed, including non existent draft-moisand Protocol Draft: Versioning and TLV numbering – Chairs ----------------------------------------------------- Woj presents on ANCP versioning strategy slide on the chairs slides. Call for comments or objections. None. Meeting participant consensus about proposed versioning strategy noted. Woj presents on TLVs slide in the chairs slides deck. Proposal is for new TLVs to be defined when conflict is noted, and conflicting TLVs made obsolete from the draft. PORT-UP TLV Type 0x02 (Access-Loop-Remote-Id) to have a new TLV # PORT-UP TLV 0x04 (DSL Line attributes) sub-TLV 0x90 (Access Loop Encapsulation) to have a new sub-TLV #. Derek Harkness questions on what causes the conflicts. Answer, typo and historical mishaps. Call for comments. None. Meeting participant consensus about proposed TLV numbering strategy noted. MEETING CLOSES