[IPv6]Re: Should known-local be a generalized mechanism?

[Kyle Rose <[e-mail]>]

On Sat, Aug 10, 2024 at 7:07 PM Michael Richardson
<[e-mail]> wrote:
>     > I still don't really understand what problem known-local is trying to
>     > solve outside of misconfigurations (e.g., ULA prefixes that leak into
>     > global DNS), but that aside, I simply don't need it in my deployments:
>     > my router advertises a default route that applies to both GUA and ULA,
>     > with the router actively rejecting with destination-unreachable any
>     > packets a client tries to send to a ULA not in the router's FIB, i.e.,
>
> What happens to your default route when you have lost all your connectivity?
> Yes, many of us our involved in significant engineering making sure that
> never happens, but in other situations, global connectivity might be less
> critical, impossible (disasters) or might even be unwanted at that time.
> Yet, local connectivity is still important.  That's why we have ULAs!
>
> Tim, I'm not even sure what use ULAs are in R&E networks. Why do you have
> them at all?  Or do you?

Tim didn't write that text. I did. (There appears to be some quoted
email mangling involved.)

A default route to my border router always exists. If it has no
connectivity to the public internet, those packets just get dropped.
But all internal connectivity continues to work just fine, including
(src,dst) pairs involving RFC 1918 space and ULAs within the given
site. The router decides whether VPN-connected ULAs or public internet
destinations are currently reachable or not, and clients simply have
to deal with the fallout when they get RESETs, ICMP
destination-unreachables, or timeouts.

Kyle