All,

With all the discussions about how to accomplish this I have one
fundamental doubt.

Let's take SR and ISIS or OSPF extensions for it (say RFC8667).

It seems that none of it works network wide if we instead of prefix
reachability start advertising summary routes from ABRs to
other areas/levels. That means no use for this there.

For MPLS LDP FECs already must be /32 so if we want to keep this at the
underlay level we have a way (and actually doing it today with ordered LDP
mode) and signal liveness network wide.

That means that the real practical use case has shrunk.

That said it may be still very useful and can be done with BGP recursion
when you have MPLS-over-UDP for your service tunneling.

Essentially what we are trying to race here with is BGP service routes
withdraws to speed up connectivity restoration time. Shouldn't we really
try to understand the issue (if there is one in the protocol which I
personally doubt) before we start investing essentially by building a
workaround in LSR sandbox ?

The ABR acting as state reflector directly to interested parties or via
their local relays (local area ABRs) sounds like a cool pub-sub option. But
that still only attempts to fix current proposals.

The fact that in my local area there are /32s does not really mean I can
reach those PEs from any remote sites. There can be fabric or LC data plane
bugs on the way, accidental ACLs,  broken SRH processors,
misconfigurations etc .. which can drop (all or worse some) of my packets
to a remote PE. Only inline data plane  telemetry and reaction to it will
allow me to choose a different path to a multihomed service. Again not a
problem for LSR WG I think.

Kind regards,
Robert
_______________________________________________
Lsr mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/lsr

Reply via email to