On 20 Mar 2010, at 15:32, Vasiliy Faronov wrote:
I have a related question, about describing negotiated documents in
RDF.
Suppose there's a web resource <http://example.org/> which has the
following representations:
- HTML in English, also available at <http://example.org/
index.en.html>
- HTML in Russian, also available at <http://example.org/
index.ru.html>
Let's say that one is a translation of the other.
What can I say about <http://example.org/> in this scenario? Is it,
for
example, a foaf:Document?
Yes.
Can it have a dcterms:title? Can it have two dcterms:title's, one
with @en and one with @ru?
Yes and yes.
Somehow it doesn't feel correct to me to say that this URI identifies
one document.
Only if you apply an outdated, “offline” notion of the term
“document”. Web documents are different from offline documents. Web
documents change over time, while offline documents typically don't.
Most importantly, web documents can have *multiple representations*,
in multiple formats and languages, while we don't usually think about
offline documents in that way.
(That's why the technical literature uses a different term,
“information resource”, instead of the term “web document”. That term
has its own problems though.)
I may want to assert different things about them, e.g.
text language, or link to the translator for the translated version.
You already have solved that problem by minting URIs for the
individual representations. You can assert language, translator and so
on for those URIs.
But
then, <http://example.org/> is the "canonical" URI for the resource,
the
one that the outside world is mostly supposed to use, so it makes a
lot
of sense to say *something* about it.
Anything that applies to all representations can be safely stated
using the canonical URI. It also makes sense to relate the generic
resource to the specific ones using hyperlinks or RDF statements.
Inside the HTML representations, this is easily done. In RDF, there
are several useful predicates in Dublin Core, e.g., dc:hasVersion and
dc:hasFormat.
Best,
Richard
Is there any best practice for this?
--
Vasiliy Faronov
