Re: Links refer, URIs don't (was Re: Fragment Identifiers and Agent Perspectives)

On Thu, Oct 13, 2011 at 8:54 AM, Henry S. Thompson <ht@inf.ed.ac.uk> wrote:
> I want to sketch an approaching to rethinking this architectural space
> which is similar to but different from Manu's.  For some
> background/motivation, interested parties are referred to a recent
> talk [1] of the same name.
>
> I think what underlies any number of suggestions about the
> introduction/prior existence of some _flexibility_ with respect to
> what URIs (with and/or without fragments) identity (e.g. the passages
> from 3986 and WebArch which have been quoted, Manu's proposal) is a
> more-or-less hidden recognition that the context of use of a URI can,
> or even must, be taken into account (alongside the media type of the
> document retriev(ed/able) from it, if a fragment is involved) when
> determining its referent.
>
> Purely 'linguistic' context of use may be sufficient (what kind of
> document does the URI occur in?  What markup, if any, identifies it as
> a URI?) in some cases.  In others, not only that, but also the nature
> of the agent (client application, human being, server code, . . .)
> that is as it were asking the question [tip-of-the-hat to Larry
> Masinter] may be needed.
>
> One distinction in terms of context which might prove useful is
> between what I'll call 'actionable' and 'referential' contexts.
>
>  <a href="[URI]">...</a> in a document of some ...html... media type, as
>  interpreted by a web browser is an actionable context for the
>  contained URI.
>
>  <[URI]> rdf:type ... in a turtle or N3 document as interpreted by an
>  RDF processor is a referential context for the contained URI.
>
> Four examples, increasingly contentious:
>
>  1) mailto:nadia@example.com -- in an actionable context this
>    identifies a mailbox, in a referential context it identifies a
>    person.  The context distinction is the only way I can make sense
>    of the discussion at [2].

in *some* referential contexts it might identify a person, per "local
policy" (your reference [2]). In RDF (according to my understanding of
how FOAF is currently used) it would not identify a person, even
though the context is what I would call referential. This is because
RDF generally tries to follow the URI scheme registrations, when they
provide guidance, as Henry Story has mentioned.

>  2) http://www.example.org/PurchaseOrder.xsd#Items -- in an actionable
>    context, e.g. <xi:include href="..."/> in another XML schema
>    document being processed by an XInclude-supporting processor, it
>    identifies an XML element, whereas in a referential context, it
>    identifies the Complex Type Definition component named 'Items' in
>    the schema corresponding to the named schema document.

Again, this makes sense, under the right assumptions about 'local
policy'. RDF has no 'local policy', and that's why this issue has come
up - people using RDF have to choose between the two, since the
language is so bland that there are no cues regarding whether the
intent is "action" or "reference". It makes perfect sense to have RDF
that talks about XML elements, but it also makes perfect sense to have
RDF that talks about Complex Type Definitions.

>  3) http://example.org/people.html#noah -- in an actionable context
>    this identifies an element in an (X)HTML document, in a
>    referential context (presuming there's some appropriate RDFa in
>    the relevant document) it identifies a person.

This is Manu's case

>  4) http://dbpedia.org/resource/Albert_Einstein -- in an actionable
>    context, this identifies an RDF graph, in a referential context it
>    identifies a person.

There are actually three cases here. In an actionable context you need
a document that serializes an RDF graph. For purposes of SPARQL
queries you probably mean an RDF graph. In *some* referential contexts
it could identify a person (the situation I wrote about earlier today:
Flickr, Talis, etc.).

(I forgot to include RDF graphs in my list of indirect identification
examples... if you run one through my information resource property
inference rule you'll see the graph is very different from its
serialization)

I think that if there are 'local policies' for indirect identification
in place then what you describe works just fine - it's similar to the
function/value namespaces in Lisp. Lisp programmers don't get confused
by that. The issue is what happens in the absence of any articulated
'local policy'.
- Is the 3986/AWWW "direct identification" idea one we want to keep?
- If so, in the above examples, what does a URI (esp. with fragid)
*directly* identify (in the AWWW sense)?
- Are URIs used in RDF supposed to *directly* identify in the AWWW
sense, or *indirectly* identify/refer as you describe above.  Or some
mix (in which case how do you tell which is which)?  That is, what is
RDF's 'local policy'?

There are many kinds of indirection, thus many kinds of "indirect
identification". The mailto: indirectly identifying a person is not
the same as the #noah URI indirectly identifying what the document
says it does, and that's different from indirect identification by
primary topic. I think you're suggesting classifying the possible
directions/indirections into two groups:

1. actionable - meaning related to retrieval of a document using the
URI, or a part (fragment) of it
2. referential - meaning determined through a relation (determined by
'local policy') such as "the person whose mailbox is..." or "the
thing, named in the document by the URI, and described in the
document" or "the primary topic of ..."

Your classification doesn't match the direct/indirect distinction of
3986 which says all fragids identify indirectly. Nor does it match the
direct/indirect classification of AWWW, which says (I think) that a
URI can identify directly but not be actionable. By carving things up
in a different way your analysis does add something interesting.

Jonathan

> ht
>
> [1] http://www.ltg.ed.ac.uk/~ht/PhilWebURIs.pdf
> [2] http://www.w3.org/TR/webarch/#indirect-identification
> --
>       Henry S. Thompson, School of Informatics, University of Edinburgh
>      10 Crichton Street, Edinburgh EH8 9AB, SCOTLAND -- (44) 131 650-4440
>                Fax: (44) 131 651-1426, e-mail: ht@inf.ed.ac.uk
>                       URL: http://www.ltg.ed.ac.uk/~ht/
>  [mail from me _always_ has a .sig like this -- mail without it is forged spam]
>

Received on Thursday, 13 October 2011 14:56:57 UTC