- From: Dan Brickley <danbri@danbri.org>
- Date: Wed, 4 Jan 2012 16:58:03 +0100
- To: Kingsley Idehen <kidehen@openlinksw.com>
- Cc: Mo McRoberts <mo.mcroberts@bbc.co.uk>, Henry Story <henry.story@bblfish.net>, public-xg-webid@w3.org
On 4 January 2012 16:07, Kingsley Idehen <kidehen@openlinksw.com> wrote: > Yes, but I would also like you to share some insights into the challenges > associated with Linked Data deployment where de-referencable URI base Names > are for all intents and purposes a luxury, irrespective of their intrinsic > prowess. > > I am trying to remind people here that Linked Data is nuance laced albeit > powerful. The fact that I have no problem deploying linked data with my > toolset doesn't make that the norm for others. I am not getting this > dimension of my message across with success here, each attempt starts of a > series of long repetitive threads. Thus, I seeking alternative voices that > also understand (from experience) the *luxury* nature of HTTP based > de-referencable Names as required by Linked Data. For me, this comes back to a discomfort with the 'religious zeal' strand of Linked Data enthusiasm that we saw, at least 2006-9ish. Don't get me wrong; on balance that enthusiasm brought us amazing things and took the whole effort to a new level. But with wider adoption looming I think we need to be a bit more tolerant and open-minded about what we count as 'Linked Data'. I have a slightly different spin on that than you but I believe there are some common motivations. Linked Data is a great name, and much more public-friendly than 'RDF' or 'Semantic Web'. It's also a very general sounding name. But since it came from a specific technical community ... those working with W3C's RDF standards, and specifically those who cared about publishing and interconnecting RDF descriptions in the Web, ... it also came to have a fairly narrow interpretation. For some, it wasn't real proper "Linked Data" unless every mentioned entity had a URI, and that URI could be fetched in the public Web, and the resulting document was also RDF, ... and so on. Omitting URIs was a cardinal sin ( hence FOAF for some wasn't "really" proper/decent/good Linked Data, since we allowed people to be described without being URI-named). And so on. For me, "Linked Data" is a great name and brand, but it is fundamentally a pun. It's a plan on words, clevery drawing attention to two very different but -ahem- linked notions of "Link" that we have in the world of RDFish Web data. 1. Links -good old hypertext references- between documents; each of which modestly and partially describes some piece of the world. 2. Links between those things (virtual or real) described in said documents; various kinds of relationship that can hold between entities. Linked Data is about using networks (graph structures and hypertext structures, cleverly combined) to describe things. In some circumstances, one or the other aspect takes primacy; and in some circumstances, one or the other side can be a bit of a burden. That's fine so long as we don't insist too heavily that all good data must fit some rigid template of being hypertext-published interconnected graphs. Linked Data is great, even if sometimes you think "ok, great that I've got triples here, but really I wanted the Excel file or the video or the GraphML". We shouldn't assume that data in its ideal form is always triples. And it's great even if you sometimes think "ok, it's nice that these descriptions are broken up into bite-sized-chunks and made accessible as Web documents ... but you know, I'd rather have a giant compressed .tar.gz without the redundancy and Web crawling sometimes....". To insist that data must *always* be expressed as RDF (or EAV if you prefer the broader/vaguer term) is stubborn; to insist that it must always be split up and published per-entity in thousands or millions of cookier-cutter documents is equally stubborn. If we aspire towards universal adoption we have to be comfortable with relaxing each of these constraints, even while showing the value that they can bring... I have no problem using the phrase "Linked Data" even if the data comes over CORBA; it's an evocative label for data that is structured in a way that embraces sharing and mixing, rather than a specific fixed recipe. For me the RDF effort is central to this, but we can do this modestly: sometimes RDF (and EAV structures in general) just isn't a great way of representing data, except as a metadata envelope. I don't think we can impose much more on the wider world... Dan
Received on Wednesday, 4 January 2012 16:27:10 UTC