Re: [quicwg/base-drafts] disable_active_migration with SPA (#3765)

[Mike Bishop <[e-mail]>]

> However, one of the reasons that you might want to use disable_active_migration is if you're bad at connection IDs, and instead need to prevent the client from changing addresses.
> This is now awkward because we can't really assume both things are true at the same time.

I think you're right that there are multiple levels of wanting to disable migration.

- There are the 4-tuple load balancers who really can't handle a change of anything in the tuple, so you'd better do your best to keep your NAT binding alive, because a rebind will toast your connection; forget moving to any other networks.
- There are 2-tuple load balancers who can handle NAT rebinds that leave you on the same IP (common, but not guaranteed); client might be okay to permit rebinding, but that's a risk if the IP changes, and you'd certainly better not change the IP yourself.
- There are servers that either will not (contracts) or should not (performance) serve you if you change to a totally different network attachment, but changes within the same network attachment (e.g. Ethernet to Wi-Fi, roaming to a different VLAN) would still be okay.  Those are probably rare enough we could abstract away to "don't change interface."

This makes me wonder how far we'd get if disable_active_migration could differentiate "don't change port" or "don't change interface."  Would we still need to give different answers based on handshake vs. SPA address?

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/quicwg/base-drafts/issues/3765#issuecomment-658204005