Meeting: IETF 89, Tuesday 4 March 2014.
Location: London Hilton Metropole, Buckingham Suite, 0900-1130
Chairs: Julien Meuric <email@example.com>, JP Vasseur <firstname.lastname@example.org>
Minutes: Jonathan Hardwick <email@example.com>
1.1. Administrivia, Agenda Bashing (chairs, 5 min) [5/150]
The WG secretary Daniel King sent apologies for his absence, owing to an unfortunate agenda clash with the vnfpool BoF.
1.2. WG Status (chairs, 15 min) 
The chairs noted that the agenda was full. The chairs reminded the WG to use the list to initiate discussion on new drafts, and to request meeting agenda slots only after doing this, and then only if necessary.
Some milestones have been missed. The chairs will update the milestone dates.
The chairs highlighted three new areas of work that are on the horizon for the PCE WG after the current milestones have been addressed:
· Extensions related to the new SPRING WG
· Time Sensitive Networking (L2 flow steering)
2. WG IDs
2.1. PCEP Extensions for GMPLS (Oscar Gonzales de Dios, 10 min) 
Oscar confirmed that the recent comments on the list have been addressed, and as a result the processing rules and syntax has been simplified. The authors believe that the document is ready for last call.
No comments from the room.
The chairs agreed to put this document in the queue to last call before the next IETF.
2.2. Domain Sequence Encoding (Dhruv Dhody, 10 min) 
Dhruv outlined two alternative options for encoding the new IRO object type, and polled the meeting for opinions.
Jon Hardwick voiced a concern that a new IRO object type may not be necessary. After a brief discussion, it was agreed to take the question to the mailing list.
2.3. Status of Stateful PCE I-Ds (Ed Crabbe/Jan Medved, 10 min) 
Ed informed the meeting that these documents have seen little change recently, and in his opinion are ready for last call
Julien said that the applicability draft still needs a bit more work, and that both should progress together. The target for this work is September 2014.
Ramon requested to include the LSP object in base draft, because it is needed for some use cases. Ed preferred to cover this in a separate draft, but allowed that this is an editorial issue that is still open for discussion.
3. Previously Discussed IDs
3.1. Secured Transport for PCEP (Diego Lopez, 10 min) 
JP requested the authors to discuss the requirements for handling DoS attacks with the transport group and to post a summary of the discussion on the PCE list.
JP noted that the name PCEPS is maybe confusing for people, as it sounds like new protocol.
There was discussion about whether the DNS section was premature, but Diego clarified that this is about using the DNS instead of manually installing certificates, not for discovery.
JP and Adrian encouraged the working group to show interest and support for this work as it is of great importance.
Adrian requested the draft authors to add a paragraph discussing the considerations for handling key rotation.
Dan York was interested in talking to the draft authors about their use of DANE. It was agreed to have this discussion outside the meeting.
JP polled the room for people that have ready the draft (a large proportion of participants) and that support adoption (roughly the same.) JP noted that there is good support for this draft and will issue a poll for adoption on the mailing list.
3.2. PCE Discovery Using DNS (Qin Wu, 10 min) 
There was a discussion about whether there was a real-world requirement for DNS discovery of PCE servers.
JP (speaking as an individual contributor) argued that auto-discovery within the domain by the IGP is not burdensome as PCE attributes are relatively static, and that the intra-domain case can be handled by static provisioning as the number of domains is not large.
Diego spoke in favour of the requirement explaining that where possible, provisioning by “human needlework” should be eliminated to reduce OPEX. Diego has a potential use case in inter data-centre connectivity.
JP polled the room to find out how many had read the draft (order 10 people).
JP encouraged the authors to start a discussion on the mailing list to show that there is support for this.
3.3. PCE Path Profiles (Santiago Alvarez, 10 min) 
Haomian Zheng asked whether PCC and PCE initiated mutually exclusive with each other, that is, whether they can they exist simultaneously for one path computation. It was agreed to take this discussion offline. Julien noted that the discussion went beyond the scope of this one draft.
Dhruv requested that the authors add a motivation section to the draft. He also requested clarification on how the path profile IDs are known / agreed by the PCC and PCE. He had detailed comments on the object format which he agreed to send to the list.
3.4. PCE-Initiated GMPLS LSPs (Zafar Ali, 10 min) 
Zafar requested WG adoption of this draft.
Julien polled the room twice, to find out (a) who had read the draft (about 10 hands) and (b) who would support adoption? (about the same hands)
Julien noted that this was not too many people, but agreed to poll the list for adoption.
3.5. Optimization of LSP DB Synchronization (Dhruv Dhody, 10 min) 
JP asked whether there were any implementations of this function.
Dhruv explained that sync avoidance is implemented by multiple parties, but incremental re-sync was to his knowledge only implemented by his company.
Dhruv requested WG adoption of this draft.
JP polled the room twice, to find out (a) who had read the draft (about 20 hands) and (b) who would support adoption? (about the same hands)
The chairs agreed to poll the mailing list for adoption.
4. New I-Ds not Discussed...
4.1. PCEP Extensions for Cross-Domain Label Distribution (Huaimo Chen, 5 min) 
No questions or comments.
4.2. PCEP Extensions for Link Bandwidth Utilization (Dhruv Dhody, 10 min) 
JP suggested that the authors search the IETF archives for discussions about the Russian Dolls model for pools of bandwidth.
There was discussion between Igor and Dhruv about whether the constraint should be based on TE advertisements. Julien noted that the TE advertisements were already being standardized by the OSPF and IS-IS WGs and that this work is necessary for PCEP to keep in step with the IGPs.
David Wood expressed a doubt that this was a worthwhile objective function to standardize, and compared it with regular CSPF.
Dieter Beller also expressed doubt that this would be required and suggested that the need for this could be avoided by proper network planning. Dhruv explained that his motivation was for the case that the network operator is unable to rely on perfect pre-planning.
Julien advised the authors to socialize this work on the list before the next IETF, before the WG considers the next steps for this draft.
4.3. PCE Support for Domain Diversity (Dhruv Dhody, 5 min) 
JP (speaking as an individual contributor) and Igor disputed that it is useful to have an OF that minimizes the number of shared domains.
Adrian (speaking as an individual contributor) remarked that if the only motivation of this work was to maximize diversity then the WG should focus first on developing an objective function that achieves this within a single domain first. (He noted that there may also be other commercial / geopolitical reasons for minimizing common transit domains.)
Dhruv stated that his intention was to request the WG to adopt this draft.
JP requested to see more discussion on the mailing list before discussing adoption.
4.4. Stateful P2MP (Dhruv Dhody, 10 min) 
Zafar requested a new capability bit for “stateful P2MP path computation”, and this was agreed by Dhruv.
Dhruv again asked the WG to consider these documents for adoption.
Julien requested to see more discussion on the mailing list before discussing adoption.
4.5. PCEP Extensions for Multiple Sources and Destinations (Dhruv Dhody, 5 min) 
There was a discussion about whether there was hard evidence to suggest that encode / decode was a bottleneck to performance, and to what extent this approach would improve performance.
It was noted that the draft changed the semantic of the request / reply such that it was not clear how to report failures in the case that some paths could be computed and others not. Dhruv accepted this but said it could be addressed by making the PCRep encoding more sophisticated.
Several contributors noted that the path-profile work would have a similar benefit in compressing the encoding of PCReq / PCRep but without the disadvantage of fundamentally changing the semantic of the request / reply.
4.6. PCEP Extensions for Resource Sharing (Haomian Zheng, 5 min) 
There was a comment that this was a useful draft/problem, and a suggestion to add new parameters that allow sharing with different BW requirement.
4.7. PCE for Software-Defined Transport Network Control (Haomian Zheng, 5 min) 
Julien deferred comments to the mailing list for lack of time, and referred the draft authors to the ABNO architecture.
5. IDs Seen Before not Discussed...
Stateful PCE for Data Plane Switchover and Balancing (Yosuke Tanaka, 5 min) [150/150]
There were no comments or questions.
Julien closed the meeting.