URI and Web Architecture Revisited

On 15/11/2007, Mark Wilkinson <markw@illuminae.com> wrote:
>
> Xiaoshu, This is a beautiful synopsis of the problem - THANK YOU for
> taking the time to write it up as well as you did!  I will be using this
> in my lectures for sure!  :-)
>
> What made me chuckle was how similar the DFDF concept is to the LSID
> concept... except that the LSID doesn't rely on any HTTP response codes to
> determine what is what (it's all explicit, since there is no way to ask a
> URI what it is but to query the metadata; and there is no concept of
> content-negotiation in LSIDs since all various representations should be
> referred to explicitly in the metadata... so there are no shortcuts to
> getting the representation that you desire)
>
> Given that we have millions (billions?) of URLs out there in the world,
> isn't it a bit optimistic to assume that they will all suddenly become
> adherent to whatever we decide here?
>
> Personally, I am inclined to place my trust in a Semantic Web where I know
> that the URIs I encounter are guarateed to have the behaviour that I
> expect.  If I can't guarantee that from a URL (and I know that I can't),
> then I can at least code my software to be more trusting of other kinds of
> URIs... and non-trusting of URLs...
>
> ...na?
>
> M

Personally I think the plain old HTTP REST proposals which include the
data type in the URL are more valuable than a philosophical statement
about the difference between information and non-information
resources. They do not need to worry about response codes because you
know what you asked for in the URL, by default getting metadata which
provides the information about the other REST url's which will return
different formats.

In the case that you actually want to use Semantic Web Enhanced HTML
pages you still have problems but they are not related to definitions
of which resource is which as the paper seems to argue. RDFa, meta
tags and link rel="alternate" seem to be valid solutions for putting
semantics inside HTML. XSLT stylesheets can be used to style RDF into
HTML as an alternative.

Do you have any scientifically relevant semantics that can't be
represented by enhancing HTML, or by using a strictly RDF directing
URL as an identifier with other information being directed to from
there? It is not hard to utilise resources if you know identifiers are
going to return RDF, or HTML pages will have link rel=alternate in
them.

Peter

Received on Wednesday, 14 November 2007 06:09:51 UTC