Re: [quicwg/base-drafts] Equivalence of preferred_address and NEW_CONNECTION_ID (#3560)

[Martin Thomson <[e-mail]>]

So I see a number of ways forward here:

1. Do nothing.  Leave this ambiguous and potentially special.

2. Make the fields in preferred_address exactly equivalent to a NEW_CONNECTION_ID with a sequence number of 1.  But allow the connection ID to be zero-length.  A zero-length connection ID provides a new stateless reset token that is bound to the new address.  (This is a weird option that would go against previous decisions to eliminate special handling for this connection ID, but I include it for the sake of completeness.)

3. As 2, except that the new stateless reset token has connection-wide scope such that either of the provided tokens can be used.  This is more consistent with established principles.

4. Prohibit the use of zero-length connection IDs in preferred_address.  That is, you can't use preferred_address without a connection ID.  That is, preferred_address effectively is a new server address plus a NEW_CONNECTION_ID frame.

5. Something else.

>From my perspective, I rank these as 4 > 3  >> 1 > 2.  But that is largely based on my bias against zero-length connection IDs (as much as it seems like that might be good for privacy, I think that privacy is better served by the use of connection IDs).  If people intend to use preferred_address to migrate to a connection-specific address, then I'm OK with 3; we already have #3563 to cover the privacy hole that arises from that.

-- 
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/3560#issuecomment-606992985