On Fri, Jul 11, 2008 at 4:25 PM, Chris Chabot <[EMAIL PROTECTED]> wrote:
> See to me it's the most interesting part, maybe you are extrapolating to > far or maybe we're not extrapolating far enough. > > Why its of particular interest to me is because we have to put the same > mechanics in the php version, and we don't have any abdera type solutions > there, so it's very important to me to try to be aware of any snafu's i may > have overlooked so far. > > I personally thought the intention of the spec was to be able to produce > the same kind of information, in, well currently 2 formats, but that might > expand into more formats in the future, so that app developers with a > certain preference (json or atom) can have the information as they want. > > I didn't see any specific mentions of supporting all of > http://www.ietf.org/rfc/rfc4287.txt and > http://www.ietf.org/rfc/rfc5023.txt so -my assumption- was that that > wasn't the goal of the spec, and hence nor of shindig. Implementing a spec in a way that doesn't comply with it seems a bad idea, but "Fully" covers a lot of grey areas. Implementing it in a way that satisfies the MUST requirements is different from satisfying all the SHOULD and MAY requirements. One way to help settle this would be to complete the converged test suite for the APIs and try both implementations against them; an advantage of AtomPub and Atom is that there are mature test suites and validators for compliance with the spec, though of course the details of the data representation that is specific to OpenSocial will not be in those. I think at least we've come close to defining where the difference of > opinion is when it comes to the 2 different java based implementations, you > feel strongly that we should have full Atom Publishing Protocol & Atom > Syndication Format support, where i think I (and probably also Cassie) just > want to support what's in the spec, without being implementing everything > from those RFC's. > > I think that does bring the discussion back to the spec list though, should > the Atom part of RESTful API be completely rfc4287 and rfc5023 compliant and > support all it's features? > > -- Chris > > > On Jul 12, 2008, at 1:02 AM, David Primmer wrote: > > I've posted quite a few long emails to this list and the spec asking >> these questions and had very few substantive replies. I also don't >> think I've been making absolute statements as you claim and I'm sorry >> I've given you that impression. There doesn't seem to be much serious >> interest in implementing rfc 5023 and 4287 and so that's why I >> suggested maybe we not have such a strict requirement on it or have it >> be so prominent in the spec. The spec never says "this is an atom >> rfc-compliant service", but instead says stuff like "use an AtomPub >> POST" or "use AtomPub optimistic concurrency" so maybe I was just >> extrapolating too far. >> Maybe to others, I'm splitting hairs, but I think there's a big >> difference between supporting atom pub and supporting some >> newly-defined restful service that has xml and json payload, but I >> don't feel it's a good use of m time writing long explanations of the >> differences. There are rfc's that do that. >> > >
