 |
 |
There's one thing at least I'd argue Linked Data gets right compared to OAI PMH: It does not go against the grain of the web (no funny mini-protocol tunneled over HTTP, no resumption token stuff that assumes stateful servers). And this point really is the most important one for the way I use Linked Data: I think of it as a uniform data access protocol. One that comes with client libraries like curl, and one that is already understood by most crawlers. So in my job - publishing linguistic databases on the web - I came to think of Linked Data as the API and the lowest level of service (no HTML, no visualization, just the raw data) which can even be supported with just plain files on a webserver. So while it may still be hard to serve Linked Data, it typically is a lot easier than serving a full blown web application as front end to your database. On Wed, Sep 4, 2013 at 4:01 PM, Eric Lease Morgan <[log in to unmask]> wrote: > On Sep 4, 2013, at 9:42 AM, Eric Lease Morgan <[log in to unmask]> wrote: > > >> I get the basic concepts of linked data. But what I don't understand is > >> why the idea has been around so long, yet there seems to be a dearth of > >> useful applications that live up to the hype. So, what I want to learn > >> about linked data is: who's using it effectively? Maybe there's lots of > >> stuff out there that I just don't know about? > > > > I've been doing some reading and evaluating in the regard to Linked Data > [0], and I think the problem is multi-diminentional: > > > And here is yet another perspective. Maybe Linked Data is really too hard > to implement. Think OAI-PMH. It was suppose to be a low barrier method for > making metadata available to the world -- an idea not dissimilar to the > ideas behind Linked Data and the Semantic Web. Heck, all you needed was > Dublin Core and the creation of various XML streams distributed by servers > who knew only a handful of commands. > > Unfortunately, few people went beyond Dublin Core and the weaknesses of > the vocabulary became extremely apparent. Just look at the OAI available > from things like ContentDM -- thin to say the least. In the end OAI was not > seen as low barrier as once thought. Low barrier for computer types, but > not necessarily so for others. From the concluding remarks in a 2006 paper > by Carl Lagoze given at JCDL: > > Metadata Aggregation and “Automated Digital Libraries”: A > Retrospective on the NSDL Experience > > Over the last three years the NSDL CI team has learned that a > seemingly modest architecture based on metadata harvesting is > surprisingly difficult to manage in a large-scale implementation. > The administrative difficulties result from a combination of > provider difficulties with OAI-PMH and Dublin Core, the > complexities in consistent handling of multiple metadata feeds > over a large number of iterations, and the limitations of > metadata quality remediation. > > http://arxiv.org/pdf/cs/0601125.pdf > > The issues with Linked Data and the Semantic Web may be similar, but does > that mean we should give it a try? > > -- > Eric Lease Morgan >