Hi Dhruv, Please see inline as [GF] Regards, Giuseppe
-----Original Message----- From: Dhruv Dhody [mailto:[email protected]] Sent: Monday, July 27, 2020 7:16 PM To: Giuseppe Fioccola <[email protected]> Cc: [email protected]; [email protected] Subject: Re: [Pce] FW: New Version Notification for draft-chen-pce-sr-policy-ifit-02.txt Hi Giuseppe, Snipping to... > (0) Can these TLVs be carried inside the LSPA object, and allow the > IFIT to be used for all paths (including SR, SRv6, RSVP-TE, PCECC, > etc). I understand that your key case is with SR-Policy but it is > better to put the TLVs in a generic object rather than in an SR > specific one. I am guessing you might be influenced by BGP SR Policy > (where it makes sense to limit it to SR-Policy as BGP is used only for > SR-Policy) but in the case of PCEP, making it generic is a better idea IMHO. > > [GF]: We can surely make it generic and define the same TLVs carried inside > the LSPA object in order to be applied for all path types, as long as they > support the relevant data plane telemetry method. The current version of the > draft already mentions the possibility to generalize the proposal but through > RFC8697: "Note that the IFIT attributes here described can also be > generalized and included as TLVs for other Association Groups". [Dhruv]: Encoding these TLVs in an ASSOCIATION object of any type would not be the best encoding option in my opinion. The association types already defined are Protection, Diversity, Policy, and the next set coming up are Bi-direction (multiple types), VN association, SR policy etc. It would be out of place to put IFIT attribute TLV in all these associations. [GF]: I see your point. Since we started to focus on SR, in the context of draft-ietf-pce-segment-routing-policy-cp, it makes sense to add IFIT attribute TLV to the SRPAG. But, looking at the association types already defined, I agree, the IFIT attribute TLVs cannot be generalized in this way. LSPA object seems to be a better place to encode and would be applicable to all path-setup-types by default. Check out RFC 8733 on how to use LSPA to encode new attributes. [GF]: Thanks for the reference. I will have a look at RFC8733. > We only considered the case of SR Policy since IOAM and Alternate Marking are > becoming mature especially for SR, looking at the relevant adopted work for > data plane telemetry (e.g. for SRv6: draft-ietf-6man-ipv6-alt-mark, > draft-ietf-ippm-ioam-ipv6-options). Anyway the possibility to generalize the > proposal can be discussed. I guess you are also suggesting to replace this > I-D with a new I-D that has a general scope (not limited to SR policy, as it > is now), correct? [Dhruv] Let's continue to discuss during the meeting tomorrow. [GF]: Sure. Thanks! Dhruv _______________________________________________ Pce mailing list [email protected] https://www.ietf.org/mailman/listinfo/pce
