W3C home > Mailing lists > Public > semantic-web@w3.org > November 2006

Re: "Hash URIs" and content negotiation

From: Pius A. Uzamere II <pius@alum.mit.edu>
Date: Wed, 8 Nov 2006 11:51:26 -0500
Message-ID: <2aec93810611080851tb5388dcy8c9f90c436d5f1af@mail.gmail.com>
To: "Alan Ruttenberg" <alanruttenberg@gmail.com>
Cc: "Xiaoshu Wang" <wangxiao@musc.edu>, danbri@danbri.org, "Richard Cyganiak" <richard@cyganiak.de>, "Semantic Web" <semantic-web@w3.org>
As I said, you can argue this either way.  Ironically, the good point you
make about URIs being for the client rather than for the content provider,
bolsters the idea of decoupling the resource from the representation.
Discoverability is an important factor to consider when constructing URIs
and developing the content negotiation strategy.  Typically, the more
arbitrary names you've got for your resources, the less discoverable the
information in your system.

I think it's kind of nice to have a URI structure like
"/users/pius/profile."  If my user agent wants Yoruba and RDF+XML, it'll try
to render it.  If there's no Yoruba version, it'll fallback to English,
RDF+XML, and so on, depending on the content negotiation strategy.  Agents
can discover the available in numerous ways, including <LINK> elements
referring to the alternate representations.  Note that this resource is the
permanent home from which a user can find all alternate representations of
the resource.  If you instead rely upon different URIs for different
translations, you run the risk of making it hard for users to discover new
translations that have been added to the resource.

In addition to this, you've got the issue of accessibility.  You want
alternate representations to be considered semantically equivalent.  I feel
like using URIs closely coupled to the represenentation based on an
arbitrary naming scheme has a bad smell for many situations.  Again, I
reiterate that this decision probably depends on the semantics of the
problem domain.


On 11/8/06, Alan Ruttenberg <alanruttenberg@gmail.com> wrote:
> On Nov 8, 2006, at 11:08 AM, Pius A. Uzamere II wrote:
> > I think that one could argue that either way.  The way I see it,
> > the resource should be decoupled from its representation.  Here,
> > the resource is the content of the document, while the
> > representation is the language in which it is written.  In the real
> > world, though, it depends on the domain.
> >
> > If you're building a document management system, in which you are
> > representing each physically printed document as a resource, you
> > could argue that it's semantically incorrect to demote various
> > translations from resources to representations, and I'd agree with
> > you.
> >
> > If, on the other hand, you have online user profiles that you are
> > dynamically translating to different formats (e.g. RDF, XHTML, and
> > different language versions thereof), it's probably more correct to
> > treat the profile as the resource and the translations as the
> > representations.
> The problem is that I don't mint URIs for my personal use. I build
> them so that other's can use them (hopefully in ways that I didn't
> expect). If I am diligent in naming different things by different
> names then someone who wants to group those into a single concept
> can. If I group different things into a different name, each person
> who can figure out how to untangle them(when that is even possible)
> will do so in a different way.
> -Alan
Received on Thursday, 9 November 2006 01:50:13 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 1 March 2016 07:41:54 UTC