Re: [hybi] deflate-stream and masking

[Bruce Atherton <[e-mail]>]

If the suggestion is to add deflate-frame as proposed in that document 
to the spec as a known extension, then that is a tune I can hum along 
to. There would be no cost to doing that since there would be no 
requirement to implement it if someone didn't want to. The only question 
is whether deflate-frame been tested enough in implementations that we 
can be confident it is going to be stable.

Removing deflate-stream from the spec as a known extension is a separate 
question. Sure, it is useless and a terrible example of how to create a 
websocket extension, but there are probably a few people that have 
implementations. Do these extensions suddenly become "unknown"? What 
would be the consequence of that?

Personally I would be fine with the removal of deflate-stream, but I 
recognize there are issues.

On 22/06/2011 2:26 AM, Ian Fette (イアンフェッティ) wrote:
> As an individual:
>
> Yoshino-san sent out a proposal a while ago that did a much better job 
> of respecting the way extensions are supposed to work. Conceptually, I 
> think what we really should have is
>
> WebSocket Frame [ Masked Data [ Compress [ Application Data ] ] ]
>
> That is, the application data is first compressed, /then/ that 
> compressed data is masked and framed as per the rest of the spec.
>
> http://www.ietf.org/id/draft-tyoshino-hybi-websocket-perframe-deflate-00.txt
>
> -Ian
>