tools for using opaque uris

Hi All,

I asked about human-opaque URIs on the list about a month ago[1], and the
opinion from responses at the time seemed to indicate that persuing a
URI scheme with little or no human-meaningful information was worth
following up on.

The advantages to me seem to be:

- No chance of user misinterpreting the meaning of the URI by reading
it

- Less information = better persistence. 
       - e.g. No desire to change the URI if the name of the thing it
       identifies changes

- I18n friendly. 
       - nobody can interpret the URI, so doesn't matter what language
       it's in. RDFS:label has better multi-language support.

- Easier to mint  
       - don't have to decide on an 'perfect' label for the URI when
       creating it.


Unfortunately the use of human-opaque URIs does have much greater
infrastructure requirements to enable efficient use by humans.

1) Most important is a URI information lookup service. Something like
URIQA[2] is absolutely essential. 

Personally I'd say that such a service is essential anyway, but the
usefulness really gets hammered home when you can't glean any
information from the URIs yourself.  We currently use an HTTP GET
against the URI and follow 303-see-other redirects to fulfill this
requirement at work.


2) Secondly, you need a way of identifying resources to users in a
form they can understand without ambiguity.

My experience has been that it's common to find multiple URIs with the
same or similar labels, so our tools currently use a combination of
label and type(label) to achieve this. It's not perfect, but it works
for us in most cases.

e.g.
       Phil Dawes (Drkw Employee)
       Phil Dawes (Person) 
    
       FxTrader (IT Register Product Entry)
       FxTrader (FX Application)


3) Thirdly you need a way of authoring free-form RDF in a
human-readable fashion.

Unfortunately I don't currently have a workable solution for this, and
this is impeding RDF as a first-class data authoring
format. (practically all of our RDF is currently second hand,
extracted from relational databases and LDAP directories).

The main problem is that most human-oriented RDF authoring formats
seem to use QNames, which doesn't work with this approach.

I've written the veudas browser to use label/type information when
displaying RDF resources, but it isn't great for entering RDF
information yet. 

Does anybody know of any tools that might support authoring rdf in
such an environment?

(and does anybody have any experience or thoughts that might be
relevant here?)

Many thanks in advance,

Phil

[1] http://lists.w3.org/Archives/Public/www-rdf-interest/2004Apr/0005.html
[2] http://sw.nokia.com/uriqa/URIQA.html

Received on Monday, 17 May 2004 15:51:09 UTC