RE: ATOM representation of search result facets

Hi Erik,

>> If one extends or overloads atom content with Opensearch or elements in a custom or no namespace,

>>you are still changing the semantics of the content. To maintain the self-describing messages constraint,

>>you need to do something about the media type, otherwise that content remains opaque, or out-of-band.

>But would that mean that you discount all the extensions to atom? Like paging? atompub? None of these are

>mentioned in the spec for the atom media type, but I would hardly characterize the use of them as opaque.

>Any generic atom processor can work the documents and get a lot of value. Those who know some

> extensions get a better experience than the rest.


The atompub extensions have their own media type: application/atom+xml;type=feed.  So things like paging are covered by that media type:

http://tools.ietf.org/html/rfc5023#section-10.1


>Take HTML and the @profile attribute. It is an extension mechanism which isn't visible in the HTTP headers; >does that make using the profile attribute less RESTful?

>In any case, my point was to move the discussion from "is it RESTful" to "is it useful".

>OpenSearch is certainly useful, so are many of the atom extensions out there. They often enable loose

> coupling in machine-to-machine scenarios, where tight coupling is the norm.


By relying on format extensions which are not communicated, the coupling is tightened, not loosened.


Regards,

Peter