Nothing new here...that's the problem :)
Let's say I was setting up an events service for musicians. Following 's ref, Portable Contacts would seem in scope for the musicians themselves. Events happen in a location, so one part of the API I'd be interested in is the address stuff. To work with that data it might be useful to use geonames too. It's events, so let me have the place stuff from eventful as well. Geo, geo and geo - with three completelydifferent APIs:
The data may be exposed but it's there as a, er, kind of glass silo, it doesn't exactly lend itself to reuse.
Not that technically it is impossible to merge the APIs, there is no reason (business, whatever) for them to sit down and merge the APIs. This has happened in network engineering (and other domains) couple of decades ago; there have been many incompatible network protocol/hardware vendors then. It takes time to recognise the value of synchronisation.
She's right, but that time part is an issue. A few years ago everyone was talking about mashups - didn't the value become apparent then? We've had a good modelling language for sync'ing data since (say) 2004 when the RDF specs came out. The data-handling tooling came along with SPARQL in 2008. RESTful good practice ideas have spread widely in the past few years, with linked data I suppose being their counterpart in the semweb world. So why are APIs still so difficult?
Ok, that's glass-half-empty from a semweb perspective. Awareness of this tech has spread. The stuff around Rich Snippets, schema.org and HTML5 microdata demonstrate that the ideas are reaching a wider audience. (Incidentally I was impressed by JeniT's diplomacy about HTML5 in her excellent presentation - but I'm going to start referring to the stuff as HubrisML :)
A personal data point: last week I checked my Twitter "followers" for the first time in maybe 6 months. Around 150 new people. I'd estimate that 100 of them had reference to the Semantic Web (or some closely associated tech) in their profiles. I follow this tech, but still I hardly recognised any of these new folks.
I suspect Mike Amundsen might have a point when he says RDF will languish until it goes hyper (i.e. gain affordances as a hypermedia type). JeniT's talk of using HTML/XML/JSON/RDF for what it's best at probably applies - so how do you bring interactivity to RDF without it looking like it's got a goat's head stuck on it's back? Research needed (high on my list). Whatever, pragmatically the linked data API goes a long way.
Anyhow (once I get my bank balance back in the black) I intend to put a lot more effort into actually using this tech to build human-facing apps. I've a few ideas on how to operate as an Indie, a core one being that taking full advantage of what the Web has to offer (i.e. using linked data etc) offers a business advantage, everything else being equal.
(any comments to the thread on G+ please)