[whatwg] BWTP for WebSocket transfer protocol

[Jonas Sicking]

On Fri, Aug 7, 2009 at 11:20 PM, Maciej Stachowiak<mjs at apple.com> wrote:
>
> Hi Greg,
>
> On Aug 7, 2009, at 10:07 PM, Greg Wilkins wrote:
>
>>
>> Again this is valuable feedback.
>>
>> That's three -0' or -1's on the look-a-like-HTTP approach.
>>
>> I'll ponder what sort of simplifications could be made
>> if the HTTP style is dropped.
>
> I'm not sure the HTTP-style framing is necessarily a minus, it's just not
> much of a plus. I think the complexity cost is from the number of features
> added. I think a possible fruitful approach might be:
>
> - Review what features BWTP adds.
> - Pare down to just the most essential ones that provide a lot of benefit at
> the protocol level relative to doing them at the application level.
> - Leave enough extensibility that other useful protocol-level features can
> be added in a future version.
>
> I'm concerned about the 1.0 version of the protocol, whatever it may be,
> being too complex. The downsides of complexity are: (a) longer
> time-to-market for the core functionality; (b) likely worse interoperability
> in the initial implementations; (c) more risk of security bugs. On the other
> hand, I would also be concerned about deploying something that didn't have
> an elegant path to future extension.

I'd also add that not deploying something in the initial version does
run some risk that implementations won't properly support extensions
once they are added. A good example of this is HTTP pipelineing which
we've had a ton of problems deploying because implementations don't
follow the spec.

/ Jonas