Re: [quicwg/base-drafts] 5-tuple routing and SPA (#3608)

[Mike Bishop <[e-mail]>]

> This brings up a related question for me about disable_active_migration. Given a server can decide not to give out any additional connection IDs, doesn't that give it a mechanism for disabling active migration?

IIRC, we had that conversation when deciding whether to add `disable_active_migration` in the first place.  The short version is that taking this approach prevents the client from rotating CIDs on the primary path; if we want the client to be able to rotate CIDs but not migrate, they need to be distinct signals.

@erickinnear, in a way you're correct that clients will have to deal with the fallout regardless of what the server says.  However, the actual use of `disable_active_migration` is likely to trigger a little more than just "don't do that" -- a client might PING more actively to prevent NAT rebindings, for example.  I think it's this hinting aspect that's actually more valuable, since as you note, clients need to be prepared for attempted migrations to fail even with servers that tacitly support them.

> Per-definition, server-preferred address is an address that is capable of receiving packets from any address.

I don't think this is necessarily true, though I agree that we're talking about smaller and smaller fractions of server space.

-- 
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/3608#issuecomment-621822010