W3C home > Mailing lists > Public > public-lod@w3.org > April 2010

Re: What would you build with a web of data?

From: Nathan <nathan@webr3.org>
Date: Fri, 09 Apr 2010 18:00:33 +0100
Message-ID: <4BBF5D31.3000809@webr3.org>
CC: Georgi Kobilarov <georgi.kobilarov@gmx.de>, 'Matthias Samwald' <samwald@gmx.at>, 'Bernard Vatant' <bernard.vatant@mondeca.com>, public-lod <public-lod@w3.org>
Nathan wrote:
> Georgi Kobilarov wrote:
>> Hi Matthias,
>>
>>>> There are applications that re-use identifiers, and there are
>>>> applications that use single, hand-picked data sources.  But let's be
>>>> honest, that's not "using the linked data cloud". So, why's that?
>>>> There must be a reason.
>>>> Which
>>>> part of the ecosystem sucks?
>>> I don't see why limiting applications to use only a selected subset of
>> data
>>> sources equals "not really using the linked data cloud". This is like
>> saying you
>>> have not _really_ used Microsoft Word unless you use every possible font
>>> effect for each document you write. 
>> Okay, how about this analogy: would you say you read the Financial Times
>> every day, if all you do is have a look at weather forecast on the last
>> page?
>>
>>> Limiting your application to certain,
>>> selected datasets seems to be the only 'no-nonsense' way of using data out
>>> there while still taking care of reliability, relevance and trust issues.
>> What I'm saying is that Linked Open Data is "links" & "open data". If we
>> aren't using both, what's the point in making all the effort?
>> Don't get me wrong, I do believe in the potential of the Web of Data. But
>> the question is: what has to happen to get us there? And part of that is to
>> realize that we aren't there yet.  I'd be happy to discuss what I think are
>> the issue. But that's kind of pointless if the overall perception is that we
>> already have the Web of Data, and just more people need to adopt it...
> 
> because the web of data isn't a huge amount of use on it's own?
> 
> Linked (open) Data is only a small part of a much bigger tech stack and
> web architecture, most of what's needed to realize the full read/write
> web of data.
> 
> IMHO those who can need to stop covering the same well trodden linked
> data ground, take the linked data / eav / rdf / ontology data model
> thing as done, and shift focus to everything else that needs done. To
> start with:
> 
> http://www.w3.org/DesignIssues/CloudStorage.html
> http://www.w3.org/DesignIssues/ReadWriteLinkedData.html
> http://www.w3.org/DesignIssues/Diff
> http://esw.w3.org/WriteWebOfData
> http://esw.w3.org/WebAccessControl
> and:
> - RDF data validation (simple xsd:type validation)
> - RESTful http cache friendly access to data (+updating etc)
> - Graph Synchronisation
> - More work on RDF Writeback / Pingback
> - FOAF+SSL for machines (+standardization of FOAF+SSL or equiv)
> - A shift of focus to publishing granular RDF documents rather than
>   datasets (the only way to make web of data SPARQL + HTTP friendly)
> - Repositioning (in peoples minds) of SPARQL to being a query interface
>   that runs client side not server side
> - Multiple delegated Link services (how many articles link to
>   dbpedia/London and how many of them does it link back to..) in order
>   to keep data manageable and http friendly resource would need to
>   delegate linking to multiple third part link services
> - Translations between common eav data models (rdf/odata/gdata etc)
> - Tools to validate data at runtime based on ontologies, and also to
>   analyse multiple graphs and find out which values are most likely to
>   be the truth / trusted.
> - Common stable Ontologies promoted as Classes to developers, and widely
>   adopted + extended.
> 
> After all of that and likely many missing pieces are done then there is
> the whole app platform thing to cover:
> - HTTP and cache friendly http clients with local caches built in to
>   most target platforms with full PUT/DELETE/GET/POST support
> - Native support for RDF and translations between eav data models at a
>   low level (right on top of the aforementioned http layer) and also
>   eav/spo data structure support in common languages (something like
>   microsoft "M")
> - Web enabled OS' and application platforms with the above integrated.
>   (chromium os and browser extensions crossed w/ silverlight/flash type
>   runtimes for ui's)
> 
> Once all that's done and the design community has grokked and handled
> the ui challenges ahead (only partially covered by pivot and the likes),
> then maybe we'll see an influx of Linked Open Data applications.
> 
> all IMHO of course.

ps: realistically there's the whole microsoft thing to keep in the back
of our minds; they have pretty much a semi-proprietary full end to end
of most of the above, from M through OData through Pivot via silverlight
and seadragon - and realistically by 2011 this will be starting to take
off in a big way; there is a chance "linked data" could miss the boat
and become nothing more than legacy data which people transform in to
odata then use on the (by then) well supported and rolled out tech stack.

pps: google are pushing in this direction too, it won't be long before
we get a big surprise from their end (gdata + openid + oauth +
gmail/buzz-additions + chromium-os + chrome + android + comparatively
unlimited resources and thousands of amazing developers + a huge
developer community)

regardless of what anybody says, these two companies will push there own
versions of what we're doing out within the next 12-18 months, with full
developer support.

please do remember I'm a huge linked data fan & have my interests firmly
planted in linked data + read/write web - just aware of the realities at
hand.

regards!
Received on Friday, 9 April 2010 17:01:15 UTC

This archive was generated by hypermail 2.3.1 : Sunday, 31 March 2013 14:24:26 UTC