[whatwg] BWTP for WebSocket transfer protocol

[Greg Wilkins]

Jonas Sicking wrote:

> I'd rather not debate about which process should be used to get to a
> good protocol. I'd rather debate concrete proposals.

Sure.

So I think I'll keep this response short and see if I can
come up with a BWTP mkII  proposal that addresses
the feedback that I've received.



> So, can you describe an application which takes advantage of this
> added value. I.e. an application that transmits JSON, and that would
> not be put at some type of disadvantage (slower, harder to implement,
> impossible to implement etc) if it couldn't use the new 0x01 frame
> type.

An application that sends just a single content type like JSON is
not going to have any advantage from that proposal, because it
can assume that messages will be text/json;charset=utf8 and
be more efficient than anything that actually transmits that.

But any application that might send multiple types and/or charsets
will benefit.   Also any application that might want to conditionally
compress or encrypt a message would benefit.

Also browser that want to share a single connection between multiple
WebSocket API instances would benefit as they could use the metadata
to organize the virtual connections without the need to alter the
payload.


cheers