Re: URIs, used in RDF, that do not have associated documentation

Hi Noah,

2012/3/28 Noah Mendelsohn <>:
> On 3/27/2012 10:44 PM, トーレ エリクソン wrote:
>>> No, the representations can have characteristics that the resource doesn't.
>> Sure, media type, length, MD5 checksum,&c.
> ...and maybe more. It's very common to have a URI that identifies some
> document (speaking informally), and to find that the representations
> include things like advertising, breadcrumbs pointing to related pages on
> the "site" etc.

Of course. It was not by intention that my examples were all at a very
low level. I agree fully that there is a lot of very important data
about the representation, in most case more than that about the
resource. Unfortunately, current semantic web practice has completely
ignored this, and technologies like RDFa displays representation meta
data (like stylesheets) as resource meta data.

> I don't think RFC 2616 is particularly clear on whether this is good
> practice or not. You could make the case that surely the resource itself
> "includes" the advertising and breadcrumbs. I think there's another
> coherent point of view, which I tend to subscribe to, which is that a
> representation may include information, including advertising, etc., that
> we don't consider to be inherent in the state of the resource. Thus, if I
> have a URI for, say, a press release from the US White House such as [1],
> and I make an RDF statement that I "like" it, that doesn't necessarily mean
> that I am commenting on all the navigation chrome on the top, or the
> "blogroll" on the right.
> I think the URI identifies just the press release if the White House says
> it does, and it identifies the page with all the chrome if they say it
> does. If they make no statements, it may be hard to tell from the outside
> which is intended, in which case RDF statements about the URI may be
> somewhat ambiguous.

I agree completely. Without an explicit definition, preferably in RDF,
you never know what the resource is.

> Nonetheless, I think it's misleading to imply that the only interesting
> things that a representation might add that aren't inherent in the state of
> the resource are plumbing-related headers like checksums and Content-type.

Again, I believe the very opposite. My way of reading the HTML
specification focuses completely at the document in the role of
representation, not resource. This will have the concequences that
even ordinary HTMP links are not links between resources, but links
from a representation to a resource. This can seem a little
unintuitive at first, but it probably works in the end.


> Noah
> [1]

Received on Wednesday, 28 March 2012 21:49:03 UTC