RE: What is necessity for SRH, and other EH, insertion/removal?

["Adrian Farrel" <[e-mail]>]

[For the avoidance of doubt: I am not speaking as an advocate, merely a seeker after truth]

>> - The destination node is not SRH-capable
>
> In that case, how can the "destination" node be considered to be part of
> the SR domain in any way?

By the definition of SR domain in 8402, I think you are right.
But it is clearly targeted through SR.
The same question applies to the end point of an MPLS LSP where PHP is used. Is that node part of the MPLS domain? It is not MPLS-capable, but it is the end point of the LSP (meaning that a packet encapsulated in an MPLS header and given a label will end up at that node through only MPLS forwarding).

> It won't even know that the outer header is an outer header.

Well, it will see the outer header as an IP-in-IP header. Or, more precisely and depending on encapsulation at the SR source, it will see the outer header as an IP header for a packet delivered to one of its addresses and will process the payload accordingly. Business as usual.

> As far as I can see, the "penultimate" node is the exit node from
> the SR domain. So why is the "destination" node
> even a thing? It's only meaningful in terms of the encapsulated
> packet, not the encapsulating packet that includes the SRH.

I am finding the term "destination node" to be problematic in most of the threads running around this topic ☹
There is "the ultimate destination of the packet" and there is "the node at which the outer header will be stripped" and there is "tunnel end point" and there is "current value of the DA field".

Ciao,
Adrian