Hi Luan, > 1. Reachability in the election. RFC 9667 §6.3 says "not reachable" nodes > aren't eligible, but doesn't say reachable how. For Area Proxy the candidates > are Inside Routers advertising the Area Leader Sub-TLV > in their L2 LSPs. Should reachable mean L1 SPF, L2 SPF, or just a > non-zero-age LSP in the L1 LSDB? My leaning is L1 SPF, since the leader > represents the L1 area.
Yes, L1 SPF. > 2. Proxy System ID derivation. §4.2 says candidates SHOULD be configured with > the same Proxy System ID, but gives no default when it isn't configured. I > currently derive it deterministically (first 6 bytes of SHA-256 over the area > tag, forcing a zero first octet to 0x01) with an explicit override taking > precedence. Is that acceptable as a fallback, or should it always be > explicitly configured? You’re welcome to concoct your own system, but other implementations rely on configuration. If you want to interoperate, that needs to be an option. > 3. Area SID flags. §4.3.2 defines F/V/L but not who sets them. I'm treating > them as the Area Leader's local config, defaulting to F=0/V=0/L=0 (IPv4, > index, global), with no requirement that Inside Nodes agree. Does that match > the intent? Yes. > 4. Who advertises the Proxy System Identifier. §4.3.1 says other candidates > MAY also advertise it, but also calls multiple identifiers a > misconfiguration. I've gone with only the elected leader advertising it, and > candidates staying silent until they win — that avoids churn and the > multiple-identifier case. Is that the intended reading? Not exacly. The prohibition is about multiple different proxy identifiers, not multiple systems advertising the same identifier. Your approach is valid. > 5. Boundary LANs. §2 says multi-access boundary circuits with multiple Inside > Edge Routers are "not supported." Should I reject that config, warn and > continue, or leave it undefined? And is using the originally configured > sysid for the DIS pseudonode on a boundary LAN a hard requirement? How you want to handle that situation is an implementation decision. If you have a boundary LAN, then you want the pseudonode id to be the area proxy ID so that it looks correct for outside systems. However, this gets very confusing if there are multiple Inside Edge Routers generating IIH’s with the same sysid, with no obvious resolution. As the situation has become unlikely in modern deployments, we chose to leave the situation out of scope. > 6. Algorithm 0. My reading is that Area Proxy should always advertise > algorithm 0 with no flooding topology (dynamic flooding disabled), since the > leader produces a Proxy LSP rather than a flooding topology. Can you confirm? Yes, unless dynamic flooding is also in use. > 7. Proxy LSP fragmentation. The spec doesn't say whether the Proxy LSP can be > fragmented. My assumption is yes. Reuse normal IS-IS fragmentation, since > §3.1's fragment-0 rule applies to the Area Proxy > TLV in a node's own L2 LSP, not to the Proxy LSP itself. Is that right? Yes, normal fragmentation rules apply. Tony
_______________________________________________ Lsr mailing list -- [email protected] To unsubscribe send an email to [email protected]
