I'm ok with changing greasing codepoints, but I'd prefer a simpler option if we're going to do that like 27 + N*31*31? Using any codepoints in the 1 or 2 byte encodings seems a bit wasteful given how valuable they are, but after that, I can't imagine anyone cares about burning a few codepoints.
On Tue, Nov 17, 2020 at 7:45 PM Lucas Pardue <[email protected]> wrote: > That said, I am not sure if we have to change the greasing scheme at this >> point. When we noticed that the ack-delay extension was colliding with the >> greasing ID, I created the PR below, that detects such misuse at compile >> time: >> >> https://github.com/h2o/quicly/pull/416 >> >> It might just be sufficient if other endpoints implement this type of >> detection scheme as well. >> >> > Or add the failure condition to the I-D tools :D > >
