Re: [quicwg/base-drafts] Stateless Reset packet sizes should not depend on the maximum connection ID length (#2869)

[Martin Thomson <[e-mail]>]

I could live with that.  That is a lot harder to write down, but I think that it might work.  I wouldn't make the calculation so complex (just say that the minimum packet size is `maxcid+19` to account for 1 type byte, 1 data byte, and 16 auth tag bytes plus the connection ID length and the one extra byte we lose for the stateless reset) and I wouldn't tie it to a particular role (stateless reset is supposed to be symmetric, it's just that one direction in particular is more likely by far).  We can leave the details to implementations.

I was concerned about clients that were also identifiable by means easier than fingerprinting.  Say someone who deployed client code in their server cluster, where connection IDs are far more likely as a client, for whom any packet (to a client or server) could be readily identified as a stateless reset by virtue of its size.  Profiling by IP address would be easy for someone to do in that case.

-- 
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/2869#issuecomment-509039192