[quicwg/base-drafts] Short header reserved bits: make available for unilateral experimentation (#2022)

[Igor Lubashev <[e-mail]>]

(Moving a discussion about this issue from https://github.com/quicwg/base-drafts/pull/2006#discussion_r234690190)

Presently, the two Reserved bits in the short header must be 0 (after removing header protection), unless both client and server agree to use them for a specific purpose.  Unilateral experimentation by the sender using these bits is intentionally disallowed.

The suggestion is to allow these reserved bits to be set at will by the sender and be AEAD protected but not masked during header protection -- just like the latency spin bit. If not in-use for a specific purpose by the sender, the bits should be randomized for greasing.

This way, the sender can use these bits in unilateral experiments (VEC, Loss bit, etc), and the middleboxes will be able to observe but not modify the bits.

There is concern that some experiment may inadvertently compromise connection privacy. Senders should be careful not to leak private information in anything they send, including IP ID fields / IPv6 Flow Labels, TTL/HopLimit fields, DS/ECN fields, packet size choices, packet timings, congestion avoidance behaviors, as well we these two Reserved short header bits.

-- 
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/2022