Re: [hybi] Payload only compression extension, again

[Patrick McManus <[e-mail]>]

compression is a proven win on the web. I don't want to support a 
protocol without it defined and the web has waited way too long for 
websockets already, so I'm not willing to support a delay in the spec 
over refining it when there is an extension mechanism defined for 
getting experience with and deploying other approaches.

individual messages will still compress fine with stream compression 
even in the presence of the rotating mask (though I support the per 
payload compression draft moving forward independently) - the rotating 
mask just damages the backreferences between different messages. This is 
effectively the same level of impact content-encodings give you in http, 
which is significant for messages of reasonable size. That isn't the 
only use case for websockets, but it is a meaningful one.

Furthermore the mask is unidirectional making the downstream data flow 
unimpacted.