- From: T.Heath <T.Heath@open.ac.uk>
- Date: Wed, 10 Oct 2007 01:19:01 +0100
- To: "Bijan Parsia" <bparsia@cs.man.ac.uk>
- Cc: "SW-forum Web" <semantic-web@w3.org>, "Linking Open Data" <linking-open-data@simile.mit.edu>
Hi Bijan, Vaguely coherent response inline. The most interesting point actually comes towards the end... > On Oct 5, 2007, at 11:10 AM, T.Heath wrote: > > > Hi Bijan, (CC Linking Open Data list), > > > > Nice. I think we could all use more prompts to stop and reflect. > > Thanks. > > > Some > > comments inline... > > > >> Now some of these may have had other factors as well (RSS > 1.0 is an > >> obvious example). But it's not clear to me that RSS 1.0 is such a > >> great idea. If we could press a button and eliminate all the other > >> flavors and Atom, or RDFize all of them, would we do so? > Would it be > >> a good idea? > > > > Yes. Speaking purely at a practical level I really appreciate being > > able > > to manipulate RSS data using a common set of RDF tools, and wish I > > could > > do this more often. Finding non-RSS1.0 feeds is always kind of > > annoying. > > I'm sure there's a stack of other reasons related to the innate > > linky-ness of RDF that sets it apart from other syndication formats. > [snip] > > Back in the day, I was writing an article series for XML.com on RSS > 1.0 driven websites. Never finished one in the series, alas. > > Let's break this down a bit. It's clear that RDF tools work for you > and you work well with them. That's immediately a biasing factor. Absolutely. In fact I probably can't answer any of your questions objectively because I'm too biased. > Indeed, when selling to some clients my folks will often point out > that RDF is *one of many* possible solutions, with different > strengths and weaknesses, and that one of the strengths it has for > projects involving us is that we're real good at and with it. But > this is true for Smalltalk guys too. Or Lisp folk. The tools work for me because I've put the effort in to learn them, but of course that doesn't mean others will do the same without a clear business need. So I think we're agreed. > But let's suppose that RDF does given you an edge in > manipulating RSS > data (for certain tasks (more on this below (and you know I'm > nesting > like this because of the Lisp reference above :)))). How > *much* of an > edge would it have to have before that edge became a reason for > people to switch? (In general, minority tech comes with prices in > terms of infrastructure: generally there's less effort thus less > mature tools and less experienced workforce. So just winning isn't > enough, you have to win by a very large margin. That or be cool > enough so people don't care about the cost.) Yeah, agreed. Right now the edge isn't great enough from a developer's perspective. Actually I believe the overall edge is large enough (it's huuuge!), we're just looking in the wrong places. This is where Linked Data comes in... More below... [sorry, needing to <snip/> lots] > To sum: I think RDF has definitely lost the wire format war in the > syndication space, and for good reasons (are they good?). They're probably not great reasons. Perhaps the loss was as much down to timing (RDF-world being ahead of the crowd) and politics?? Maybe RSS is simple enough and not linky enough that it doesn't need anything other than vanilla XML? (No, I don't mean that! There are so many opportunities to do RSS feeds properly, Linked Data style. <dc:creator>tomheath</dc:creator> in my delicious RSS is such a joke.) > At the very > least, it doesn't provide enough of a value add and has significant > downsides. Do you disagree? (Your reply indicates so, but > because you care about the backend. Doesn't GRDDL, or even, y'know, just > transforming, do the job for you?) I agree about the lack of value add; the significance of the downsides depends who you ask. At the end of the day there are many ways to skin a cat, and people will choose the approach that suits them. These kind of issues are not crucial; demonstrating the RDF sweet spot has got to be the priority. <snip/> > You might find this post interesting as well: > > <http://clarkparsia.com/weblog/2007/03/19/why-not-sparql-for-dblp/> > > In short, I found it much, much easier to get my data linked when I > didn't go through linked data, so to speak. I think you've inadvertently hit the nail on the head. From what I read in your post you didn't actually get your data linked at all. Exhibit is really cool, but it's only really on the Web, it's not in/of the Web. Where are the links to other data? The links between data is where RDF adds the value. This is the point we need to demonstrate to people. (I know I don't need to tell you this, but just for the record... :) Why can't I follow links in your Exhibit to, say, Chris Halascheck, and then browse off around the Web looking at the rest of *his* publications (not just those he's co-authored with you), his holiday snaps, what he had for dinner last night etc etc? That's where RDF comes in, and no Web2.0 mashup will scale to this kind of scenario. The reason why people get so excited about the Linked Data "movement" is that finally we have data sets that are large enough and interlinked enough, and backed with a growing amount of infrastructure, that we can start to demonstrate these capabilities. Exhibit is a great presentation layer (watch this space for some nice ISWC2007-related Exhibits coming up), as are most Web2.0 mashups, but they're for humans only, and don't address the data layer. Phew, excitable rant over :) I dunno if this moves us any further on, but its certainly been good (and hard) thinking about it. Thanks, Tom. P.S. For the record, creating this <http://kmi.open.ac.uk/people/tom/tmp/dblp.nt> from the Hannover DBLP took about 3 mins from opening an editor to getting the nt file uploaded. Of course it's not one to show your Grandma, but a beautiful publications page (with all linked RDF data still in place behind the scenes) is just a SELECT query and an XSL transform away ;)
Received on Wednesday, 10 October 2007 00:19:23 UTC