Re: p2: Expectation extensions

[Mark Nottingham <[e-mail]>]

On 23/04/2013, at 11:57 PM, Amos Jeffries <[e-mail]> wrote:

>> Given how many problems we already have with Expect, should we consider disallowing further extensions here, and removing this requirement?
> 
> So whats gained by making it an expectation if the expectation is ignored? nothing.
> 
> Removing it also removes the fail-closed property of Expect:. I know the property is a great annoyance to new featrue rollout. But it does offer the concrete assurance that what is expected is supported which is quite useful when designing security related extensions. We have the Prefer header coming up to provide the expectation negotiation with fail-open semantics.

Disallowing future extensions recognises that interop on this feature is low -- i.e., that deploying an extension here isn't likely to work well. I don't see much in the way of "concrete assurances" through its use at all, as a protocol designer; only a lot of nasty problems.


--
Mark Nottingham   http://www.mnot.net/