Fri, Jul 08, 2022 at 05:13:37PM +0200, Robert Raszuk: > > I do not understand Robert's issues with this community. > > Let me say that I like it from operational perspective. > > However from routing perspective I have doubts on how this community is > going to be originated and how it is going to be used. >
I do not preceive these concerns to be unique to this community. rfc7999 (BLACKHOLE) is of little difference, for eg. The draft and comments on the WG ML have noted a few uses, besides it being informational. > Example 1: Hyperscaler has global POP footprint and is offering anycast > services for its VPCs. So lot's of prefixes are going to be marked as such. > Is the expectations that ISPs will/shall use it ? maybe; that is their decision, not ours nor the originating AS's. They might ignore BLACKHOLE, GRACEFUL SHUT, or no_export too. They might choose to ignore (or not) it only from Hyperscaler. They might have a specific agreement with Hyperscalar to observe the community. > Example 2: Enterprise is injecting same prefix from different locations. It > could be anycast or not. Or maybe as mentioned one host route within such > /24 is true anycast. Is ISP suppose to alter their local pref according to > such marking ? Even if there is no really ANYCAST service behind at all ? They might; both are their decision. They might not observe the community for prefixes shorther than /24 (/64). They might just not ignore (use) the MEDs received. They might not observe it for prefixes with an AS_PATH greater than 1 AS. They might advertise the more specific with no-export. > Example 3: How do we detect misuse and accidental or purpose marking by > malicious guys in the middle ? you know that is not possible today. You can choose to craft your policy to limit from whom it is accepted or have other constraints, just as with any other community. Seems like we've been over this for other communities. Besides Jeff's requested changes (your Ex 2 & 3), perhaps you just want prefix lengths addressed? _______________________________________________ GROW mailing list GROW@ietf.org https://www.ietf.org/mailman/listinfo/grow
