- From: Dan Brickley <danbri@danbri.org>
- Date: Mon, 1 Nov 2010 11:41:43 +0100
- To: Richard Light <richard@light.demon.co.uk>
- Cc: Karen Coyle <kcoyle@kcoyle.net>, public-lld@w3.org
On Mon, Nov 1, 2010 at 11:29 AM, Richard Light <richard@light.demon.co.uk> wrote: > In message <AANLkTi=cNMx-aFdp7+_ed_hCPxJ5OeshKAjJ2ePrz4GG@mail.gmail.com>, > Dan Brickley <danbri@danbri.org> writes >> >> The original Linked Data note at >> http://www.w3.org/DesignIssues/LinkedData.html was a gentle critique >> of this idiom, basically saying "give the people http URIs too!". And >> we did! > > With the emphasis on "too": a LD URI is simply a peg on which you can easily > hang a bunch of assertions. Without those assertions (a) it's not giving > you anything useful and (b) you have no means of deciding whether "your" > Richard Light URI refers to the same person as "my" Richard Light URI. Yes, exactly. "Belt and braces". The more information we have, the better. I like the 'peg' metaphor. Bnode-heavy RDF lacks those pegs, which makes it harder for others to overlay related info by publishing RDF elsewhere. >> I still have some concerns about the practice of declaring URIs for >> other people, especially living (non-technical!) ones, unless they are >> given some clear 'right of reply' if the associated descriptions are >> inaccurate. > > Well, the source of a published URI is clear through the domain registration > system, so surely the same remedies are available as for any other > publication medium? Most end users don't understand the difference between the concepts of 'browser' and 'search engine'. So they're not going to understand domain names any time soon. I agree the information is trackable if you're an expert. The trouble is that RDF is designed to flow around the network and get re-used. Take for example the page http://www.bbc.co.uk/music/artists/0ca53fff-3b07-49eb-bcb9-bbe84f1ec768 (or append .rdf to that URL to get RDF/XML). The RDF there came from MusicBrainz originally. This stuff flows around the network in ways that aren't always easy to track. The pre-LOD notion of de-referencing a URI had some sense that you were getting an 'authoritative description' of the thing by fetching it, ... asking it for a self-description. When the URI is assigned by one party about another, that notion of self-description goes away. That's my continuing mild discomfort with the idea that it's OK to propagate "http://danbri.org/example.rdf#richard_light is a URI for Richard Light", since it might be mistaken for Richard Light's 'official' self-description. At the moment most LOD RDF is published in good faith; if it ever goes truly mainstream, that will change and it'll be spammed and malware'd like the rest of the Web. So we need to start thinking now about how to distinguish these different notions of authority. Which hopefully brings this mail back on topic, since authority control and being good-faith public sector information authorities is something the library world cares about... cheers, Dan > Also, I don't see a scenario occurring where everyone self-publishes their > "official" URI and associated properties, so I think we are left with this > situation and just have to deal with it. > >> But in the general case there's no doubt that the Linked >> Data emphasis on URIs everywhere was a leap forward; data merging >> based on identifying properties is a huge pain. If we can get a URI >> for each node, a lot of things become simpler. > > At some stage the tedious "comparing of properties" task still crops up, > e.g. when deciding whether two URIs reference the same concept. > > Richard > -- > Richard Light >
Received on Monday, 1 November 2010 10:42:16 UTC