Re: Interaction model vs data model

hello pierre-antoine.

On 2013-01-24 19:12 , "Pierre-Antoine Champin"
<pierre-antoine.champin@liris.cnrs.fr> wrote:
>reading your answer above, as well as your answer to Andy's previous
>question in this thread, I think it would indeed be useful in this group
>to distinguish between *links* (or RDF-links) that merely represent a
>relationship between two entities, and *hyperlinks* (or REST-links) that
>describe potential interactions.

absolutely. actually, when you look at REST as one variation of the
general hypermedia approach, then being able to identify navigable
hyperlinks is essential for the style. different representation
metalanguages have developed their own standards or design patterns for
that, but there's a reason why they were all doing it. guided by a
protocol, clients must have the ability to selectively follow hyperlinks
to reach their application goals.

>Of course, one could argue that, according to the Follow Your Nose
>heuristics in linked data, every RDF-link is an incentive to GET the URI
>it links to, hence a hyperlink... In fact, from this point of view, any
>URI identified as such is a hyperllnk... This over-generalization does
>not seem very useful, though.

no it's not very useful, because that would only cover GET requests, and
would not guide clients through an application flow.

you could see this like throwing in random XInclude into the mix of an
XML-based protocol: it may be useful for servers to have the capability to
fragment responses on demand, and then clients are guided by the (overlaid
and very simple) XInclude protocol when they want to retrieve parts of a
resource that haven't been included. but for the actual application goal,
the essential hyperlinks still are the ones that represent the application
protocol, and where interactions with those links are governed by the
protocol ("POST something here, expect a certain state change/result").

cheers,

dret.

Received on Friday, 25 January 2013 15:16:11 UTC