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

[ianswett <[e-mail]>]

The only rule is that the CID can't become 0 length, so why is it invalid?  If the middlebox can't reliably identify the packet as invalid, it's all a matter of statistics and how much of an anomaly it is.  Sure, maybe it's an anomaly for CIDs to go from 18 to 1, but I don't think that's the most likely anomaly on the internet(see below).

I think the existing rule is actively harmful in multiple ways, one of which is a possible escalation of the minimum client packet size mentioned above.

The more problematic issue is that the goal is to have stateless reset not stick out and be clearly identifiable.  If so, we want the widest selection of sizes possible.  Given servers are going to want to send stateless resets smaller than the client's packet, a fixed size is going to stick out and the larger the minimum fixed size, the more it's going to stick out.  In particular, 40 is enough larger than the typical 'small' server to client packet(ie: ACK) that it's relatively unlikely on a typical connection.  If 90%+ of 40 byte payloads are stateless rejects, then we have failed at our goal and I think that's where we're at today.

To come back to my previous suggestion, why can't the client enforce this?


-- 
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-507909237