W3C home > Mailing lists > Public > semantic-web@w3.org > January 2011

Re: Spring cleaning

From: Paul Gearon <gearon@ieee.org>
Date: Tue, 4 Jan 2011 11:20:39 -0500
Message-ID: <AANLkTikmzzNQExG_Oa1JQovm5QHzKGLuOpWJnTVFWv08@mail.gmail.com>
To: nathan@webr3.org
Cc: Semantic Web <semantic-web@w3.org>
On Mon, Dec 27, 2010 at 9:51 AM, Nathan <nathan@webr3.org> wrote:
> Hi All,
> There are a few little bit's I've not really seen mentioned that it may be
> good to see addressed in 2011.
> As stupid as this may sound, there isn't actually any dereferenceability
> built in to RDF or "linked data", in RDF the URIs are all just logical
> constants and it's on a totally different level to the web, the RDF
> serializations make no mention of dereferencing URIs, and linked data isn't
> a protocol, thus tbh, the whole linked data thing is built purely on out of
> band knowledge between us, with /nothing/ to indicate to any machine that
> you can or can't dereference any URI mentioned in RDF.

You'll find that there are a lot of people who see this as a desirable
property. Of course, there are others who agree that everything should
be dereferenced as well. But you won't achieve consensus on this,
therefore it won't make it to standardization.

It has become best practice to use URLs (and not URNs) and to make
these dereferencable, thereby enabling linked data. That's probably as
good as we'll get for a while.

Personally, while I like to use dereferencable URIs where I can, there
are occasions where I need to mint URIs for large numbers of
individuals, and it's not practical to have these things all
dereference to individual resources. There are approaches to mitigate
some of this, but my point is that it's not *always* useful or
efficient to make URLs dereferencable, and a lot of people will want
to defend that position. The "Best Practice" approach means that it's
not compulsory, but you should have a good reason if you don't want to
do it.

> In addition to this, most of the RDF serializations aren't IANA registered
> either, other than RDF-XML, may not sound like much but it does have big
> side effects, as a quick example, if you take a turtle, n3 or rdf-json file
> and expose it via a web server, there's no mime type, the possible file
> extensions aren't recognized, and there isn't any way to conneg - even if
> you enable content negotiation, servers and clients /can't/ negotiate
> because the mime type isn't there and isn't known.

As Antoine said, there is a proposal to standardize JSON and Turtle
representations of RDF (Turtle being that part of N3 that is relevant
to RDF). A lot of us agree that this is important.

Paul Gearon
Received on Tuesday, 4 January 2011 16:24:50 UTC

This archive was generated by hypermail 2.4.0 : Thursday, 24 March 2022 20:41:25 UTC