While mulling over the UA changes in bug 588909, I got to thinking... Could we leverage SPDY as a way to correct some of the mistakes of HTTP from the past? 3 examples come to mind: 1) User-Agent headaches. Almost every time we touch this there's a flamewar and something breaks, and I don't think _anyone_ is happy with the current state of affairs. ("Mozilla/5.0"? Really?). 2) WWW-Authenticate problems with unicode. See bug 41489 for details. 3) "Referer" [sic] header misspelling. Probably others that don't come to mind at the moment. Basic idea would be to require that sites supporting SPDY must also support a few minor tweaks to the HTTP bits running over it, such that any site supporting SPDY can be presumed to also support the new behaviors.
maybe http/2 is the better scope for this kind of thing. I'd expect that to include spdy (or at least spdy-like) mechanisms but it has the opportunity to address more than the transport-binding focus of spdy.
spdy and h2 became vectors for transports rather than http semantics