Re: [quicwg/base-drafts] Connection ID Garbage Collection (#1729)

[Martin Thomson <[e-mail]>]

I have a better question: Why wasn't Christian involved with the design team work?

> [...] I would like you to stop sending anything to the CID=X"

The design team concluded that this wasn't a requirement we would address.  If you advertise a connection ID, then you have to be prepared to accept packets with that as a destination for the lifetime of the connection.  That doesn't mean that you have to sit quietly as your peer provides perfect linkability, you can definitely kill a connection if the peer acts irresponsibly, but we concluded that influencing the choice of connection ID from a peer would add too much complication.

> Does this reset cause the entire connection to crash?

Yes.  The expectation is that you can't remove any knowledge of an old connection ID until after some time has passed, even after receiving an indication that your peer has stopped using that value.  In the rare event that you spend that time and you still receive a delayed packet, then yes, you will send a stateless reset.  The hope here is that the peer has forgotten that particular stateless reset token.  That's why we have a recommendation for a timer.  We probably need text in the PR explaining that.  (@erickinnear, if you don't feel inclined to add that, please open an issue.)


-- 
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/1729#issuecomment-420128333