Re: [quicwg/base-drafts] Why are we bothering to number CIDs? (#2159)

[ekr <[e-mail]>]

I'm not persuaded by this. There are two cases:

1. An erroneous duplicate NCID.
2. Duplicated packets in the network.

If the other side is broken, there's no guarantee of consistency between
the sides in any case. So, we're only talking about the second case. And I
don't think it's particularly difficult to pick a timeout here, because
we're talking about the race between RETIRE_CONNECTION_ID and the packets
duplicate in the network, so a timeout on the order of 2MSL is basically
fine.





On Thu, Dec 13, 2018 at 2:33 PM Mike Bishop <[e-mail]>
wrote:

> Because you're not realistically going to retain a total history of all
> CIDs you've ever seen. You MAY treat it as an error if you notice that one
> recurs, so that servers are disincented from doing so, but you're not
> obliged to check that over the whole connection lifetime. However, without
> maintaining that history for a sufficiently long time, you then can't
> differentiate for certain between a retransmission of an old NCID frame and
> a new issuance of the same CID (an error, but one you might miss).
> Therefore, you could wind up in a state where the peers disagree about
> whether a CID is still live.
>
> The options were a requirement to keep history for a certain length of
> time (and hope that packets don't get delayed longer than whatever
> arbitrary timeout we pick) or using sequence numbers and allowing tracking
> with a similar data structure as that which tracks whether Stream IDs are
> not-yet-open, open, or closed.
>
> —
> You are receiving this because you authored the thread.
> Reply to this email directly, view it on GitHub
> <https://github.com/quicwg/base-drafts/issues/2159#issuecomment-447145188>,
> or mute the thread
> <https://github.com/notifications/unsubscribe-auth/ABD1oceRNi0ITMvVBtKHr-pt0lwaBKKxks5u4tXMgaJpZM4ZSa1C>
> .
>


-- 
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/2159#issuecomment-447149186