Re: [IPv6] Second Working Group Last Call for <draft-ietf-6man-rfc6724-update>

[Jared Mauch <[e-mail]>]

> On Apr 15, 2024, at 2:32 PM, Kyle Rose <krose=[e-mail]> wrote:
> 
> On Mon, Apr 15, 2024 at 2:25 PM Ted Lemon <[e-mail]> wrote:
> Okay, let's say your case is "natural." In your setup, you'd need to make fc00::/6 a "known-local" ULA, and then you would get the behavior you want, right? And you could do this by always publishing an fc00::/6 route in all your default routers (or a subset, if that makes more sense).fc00::/6 is not a "ULA prefix" in the strict sense implied by the generation procedure, which has them at 48 bits with 40 bits of entropy. If the proposal is for known-local to support arbitrary prefixes within the ULA address space, then yes, I think that would work. It's viscerally unsatisfying to treat them so differently from GUA (i.e., requiring an extra route in addition to the default route that should cover it) simply because we want the shoehorn reachability into address selection, but as long as I can achieve what I want with O(1) work, I'm not going to complain too much.

Yes, this is similar to creating what we in the operator space might call a hold-down route, eg: trying to make one for each rfc1918 type space, including the lab, docs space, etc to prevent traffic, so while a ::/0 or 0/0 would cover all that space, one can create a number of hold-down routes to prevent traffic from leaking but folks often forget to do this.

- Jared