W3C home > Mailing lists > Public > semantic-web@w3.org > November 2006

Re: ANN: D2R Server publishing the DBLP Bibliography as Linked Data (Semantic Web grows 10%)

From: Dan Brickley <danbri@danbri.org>
Date: Fri, 10 Nov 2006 15:46:13 -0300
Message-ID: <4554C8F5.8090007@danbri.org>
To: Damian Steer <pldms@mac.com>
Cc: semantic-web@w3.org, timbl@w3.org

Damian Steer wrote:
> On 5 Nov 2006, at 16:51, Tim Berners-Lee wrote:
>> We discussed it non #swig before and after
>>  http://chatlogs.planetrdf.com/swig/2006-11-04#T21-59-35
>> I liked your proposal to change the seeAlso to moreData, now 
>> implemented -- thanks you!
> I'm not sure I understand the point of moreData. It's supposed to be 
> weaker than the (weak already) seeAlso, to the point where it's a 
> relation which doesn't relate subject to object? ("Further information 
> that may or may not be related to the subject resource")

Yep, I think that summarises my discomfort too. RDFCore had the choice 
to make seeAlso more specific, but we went with the loose version. There 
is no restriction on syntax, dataset size, availability (ACLs etc). Or 
even that the referenced data encodes RDF statements about the same 
thing (let alone that it use the same URI for it).

If dataset size is an issue, ... that can be probed by HTTP HEAD.

It seems the motivation is a mix of 2 concerns: file size and relevance.

eg. If I know you live in Bristol (...and I have my spies) then 
downloading a 3k graph giving some info on Bristol is a useful little 
dataset to "see Also" if I'm doing some RDF-based stalking of you. But 
downloading everything-ever-said-about-bristol.rdf too. Similarly, ... 
if I know you're working at University of Bristol, ... I might be 
interested to know what papers you have co-authored with your 
colleagues. Now if there is a damiansteer-bibiography.rdf file, that 
seems a great thing to "seeAlso" to. If there is a 
ilrtstaff-bibliography.rdf file, ... perhaps I'd seeAlso that as well 
(testing the robustness of software if it turns out to be rather large).
But an everything-ever-written-by-people-at-bristoluni.rdf or a 
complete-dump-of-the-compsci-literature.rdf document would have a less 
attractive bulk-vs-relevance profile.

And of course relevance and utility is pretty task-specific, as well as 
qualified by the resources available. If I'm doing something on a mobile 
phone over expensive, gappy and unreliable networking, the cost of 
de-referencing some seeAlso'd data might be too high.

This situation to my mind should encourage us to find better ways of 
characteristing RDF documents, datasets, databases.... and the various 
ways they can be subsetted into useful chunks. Perhaps some notion of 
RDF "document types" couched in terms of template instances, or the 
SPARQL queries that the doctypes might satisfy.

Overloading or replacing seeAlso strikes me as not likely to come close 
to what we need here. In particular, as you say, if the referenced 
document isn't very pertinent, ... using a typed relation to describe it 
seems inelegant and un-necessarily constraining.



> If you want to mention a resource why not just mention it (:foo a 
> rdfs:Resource)? Or :foo a :InterestingPlaceToLook?
Received on Friday, 10 November 2006 18:46:40 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 1 March 2016 07:41:54 UTC