Re: [v6ops] Scope of Unique Local IPv6 Unicast Addresses (Fwd: New Version Notification for draft-gont-6man-ipv6-ula-scope-00.txt)

[Ted Lemon <[e-mail]>]

On Jan 7, 2021, at 11:03 AM, Michael Richardson <[e-mail]> wrote:
> No, it has to get into the wire format, and it has to be cachable on the
> client, and it has to evaluated on the client as close to the application as
> possible.
> 
> Why? because caching is good, and throwing away your cache each time you have
> a link up/down is dumb in my opinion.

I don’t think this would help, because you’re assuming that the data would be correct, and that the host would be able to tell which view to use. This is Really Hard. I don’t think that relying on this operationally could ever be a good tradeoff to save a few round trips to the nearest cache.

Since we don’t have a way to reliably identify links, and for that matter to tell which link is in which zone, we can’t do the efficient thing you are talking about. In practice, all resolvers that I know of (admittedly, this is just one, mDNSResponder) do in fact throw away the cache whenever the link changes, for exactly the reason we are talking about: just because the info I got from the resolver was valid on the link I was previously on, doesn’t mean it’s valid on the link I’m currently on.

Remember, just because something is efficient in principle, doesn’t mean that it’s efficient in practice. :)