Re: wildcard resource representation

This discussion seems to have forked but I'll try and shed light on this 
sub-query.

Garret Wilson wrote:
[..]

>>
>> [] a wdr:ResourceSet;
>>    wdr:includeSchemes "mailto";
> 
> This is very interesting. Two questions. First, how stable is this? 
> Should the namespace http://www.w3.org/2007/05/powder# change?

The namespace is as stable as any public draft. That is, it would be 
wrong to assume that the work we're doing will continue smoothly along 
the track to Recommendation without changes, we don't foresee a change 
in the namespace itself. Further, such review as there has been of the 
work to date has not challenged the general principles we're working on. 
That may change, of course, but for now we're confident in the general 
approach. There'll be a new draft of the doc [1] next month.

> 
> Secondly, how do I specify "all resources"? (In my case, some resources 
> may use mailto: URIs, others may use UUID URIs.) Am I stuck with the 
> following, or is there a shorthand representation?
> 
> <wdr:ResourceSet>
>  <wdr:includeUriPattern>*</wdr:includeUriPattern>
> </wdr:ResourceSet>

Our work is, of course, driven by its use cases. Ours are centred on 
things like trustmarks, standards compliance and, potentially, 
licensing. In all those scenarios you need to be describing a defined 
set of resources, typically those collected in a Web site. Therefore we 
believe some restriction on what is being described is necessary and we 
define <wdr:ResourceSet /> as the empty set. That said, your example 
raises an issue we've not discussed, i.e. what if the RS definition is 
tantamount to the universal set. Hmmmm...

A thought, you can define Resource Sets as unions of other Resource Sets 
so if you wanted 'everything with a mailto: scheme or a UUID' that 
shouldn't be hard. And the whole thing is extensible anyway so defining 
your RS to include what you want and not everything else should be 
possible (or we've goofed on the whole thing!!)

The basic model to remember is that you start with a known 'candidate 
resource' and then see if it matches the properties defined in the RS. 
If it does, it's an element of the RS. Without the starting point of a 
known resource, Resource Set definitions don't work (this is all set out 
in the doc).

However... there's more to POWDER than Resource Sets. The Description 
part is potentially more generic. I'm not sure if this helps but we are 
defining a Class called Descriptors - and this is where we step outside 
the RDF model - the properties of which are applied to whatever the 
subject of triple may be.

So, for example, the basic nuts and bolts are already in place on 
my.opera.com whereby you can tag anything you like and then link that 
tag to a pre-defined Descriptors class. My own area of child protection 
leads me to define things like 'xxx' and 'gambling'. You can call it 
what you like but if you link your tag to 
http://repository.icra.org/generic#xxx the thing you're tagging will 
also have a description that is anchored in something with defined 
semantics. The crucial point being that the properties of the 
Descriptors do not describe the tag, but the resource that the tag 
describes.

Hope this helps.

Phil.

[1] http://www.w3.org/TR/2007/WD-powder-grouping-20070709/

Received on Thursday, 2 August 2007 08:36:46 UTC