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

[Kazuho Oku <[e-mail]>]

@igorlord 
> However, there are deployments where servers still have a preferred address, but client migration using either server address is not likely to work well,

I think the question is how such an endpoint running the preferred address distinguish between the connections that migrate to that preferred address?

When a client migrates to the preferred address, the client-side tuple is allowed to change. Even if the client implementation continues to use the same port, the tuple could still change due to a NAT in the middle. Assuming that is the case, a server providing SPA is, in practice, required to be capable of using the connection ID to determine the connection. 

To paraphrase, the requirement for supporting SPA is equivalent to supporting migration. Both require CID-based routing. Hence no reason to have a knob to disable migration when SPA is being used.

> For example, a server is a backend in front of a load balancer (and hence wants to use a preferred address that does not go via the load balancer for improved performance and reliability). However, the server is in a network location that is unlikely to be accessible or be a logical choice when changing to a different network attachment provider.

I wonder if such deployment would have worked reliably even before draft-28. How can a server be sure that the client would not connect through a different network attachment provider? As stated above, it is my understanding that a client is allowed (and have to be allowed) to use any source tuple when migrating to SPA.

-- 
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-644263033