Danny Ayers schrieb:
Spontaneously I don't see the motivation for x:Resource,
rdf:Resource instead?
everything is an rdfs:resource, but I guess we know more, isn't
everything in a blogroll necessarily an infoatom:InformationResource or
a foaf:Document?
and I'm
wondering if rdf:type couldn't be used instead of x:format
:
I'd wondered about that - having a class :Rss1.0Feed, :AtomFeed etc.
Still not really sure how you reconcile the resource/representation
thing though: what if the instance of :AtomFeed also has a HTML
representation?
An :AtomFeed is a resource that *can* be dereferenced as AtomFeed, so:
http://atomowl.org/ a :RDFFeed, :XHTMLDoc, DAV:Collection.
But I'm unsure about the requirements:
- Is blogroll / directory ordered?
Not sure, I don't think there's any kind of consistent pattern out
there, so I'd be inclined to say no (unless an ordering was expressed
some other way).
I think using a list (as with parseType="Collection") can be dropped
then, it produces a much more complicated graph than having a :member
(or :resource) property for every child.
- Should the items be allowed to have titles that are specific to their
function in the directory, i.e. should it allow to reference to a
resource as "my best friend's blog" without asserting that this is the
title of that resource?
Hmm, tricky. The person that creates the list is usually pretty
important. But maybe that could be handled on top (if required) of a
title kind of property.
Hmm, either I'm not understanding you answer or the other way ...
Should it be possible that someone references your blog as
"Mediterranean information flood" without asserting that "Mediterranean
information flood" is the dc:title of your blog. "Mediterranean
information flood" is something like a source specific link caption.
In this case a construct could look like this:
<member>
<Link>
<dc:title>Mediterranean information flood</dc:title>
<target>
<Blog rdf:about="http://dannyayers.com/";>
<dc:title>raw</dc:title>
</Blog>
</target>
</Link>
</member>
Significantly more complicated, but probably the easiest way, if "source
specific link captions" are a requirement.
I was doing something a bit similar recently when implementing webdav
support in KnoBot, the vocabulary is minimalistic: Resource (again),
Collection, member, no ordering, no role-titles.
Ah, webdav, why do I always forget webdav..? ;-)
How did you end up doing it in KnoBot?
For now it's in the CVS version only, an yes atomowl.org is running with
it, so you may enter webdav://atomowl.org/ in a Konqueror windows, or
http://atomowl.org/ in the open office file dialog. For now it just maps
a couple of rdf-properties to webdav properties so it neither supports
retrieval of all properties, nor adding arbitrary properties - nor move,
nor copy, nor delete, nor does it throw the mandatory exception when
creating a resource in an inexistent container - but it seems pretty
usable nevertheless. The biggest issue implementing it, was that most
clients seem to regard a trailing slash of an URL as some kind of
baroque decoration that can arbitrarily be removed or added...
reto
