Re: A proposal involving my original reason for commenting on httpRange-14

On 4/3/12 9:04 AM, Harry Halpin wrote:
> On Tue, Apr 3, 2012 at 2:26 PM, Ed Summers<ehs@pobox.com>  wrote:
>> On Mon, Apr 2, 2012 at 7:25 AM, Kingsley Idehen<kidehen@openlinksw.com>  wrote:
>>> 303 hasn't stopped facebook adding 850 million+ Linked Data profiles to the
>>> Web. We still talk about Linked Data as though its just starting when in
>>> fact its already reached critical mass etc.. DBpedia, LOD Cloud etc..
>>> combined with Facebook and Schema.org == way beyond critical mass :-)
>> Perhaps this is where you were going, but I think it's important to
>> point out that Facebook and publishers of Open Graph data ignore
>> httpRange-14 entirely, which is probably why it is working. For
>> example:
>>
>>     http://www.imdb.com/title/tt1182345/
>>
>> identifies a movie:
>>
>>     https://developers.facebook.com/tools/debug/og/object?q=http%3A%2F%2Fwww.imdb.com%2Ftitle%2Ftt1182345%2F
>>
>> Thanks for your email Harry!
> Precisely. Indeed, while no-one *forces* you to use 303 redirection,
> you currently have to use it to count as Linked Data. Which leaves
> Linked Data to the true believers like Kingsley, while lots of people
> probably just want to mint RDF data using APIs etc.

You don't have to use hashless URIs to publish Linked Data. As I am sure 
you know, I am less a believer in RDF (solely) than I am in the powerful 
idea of Web scale structured data enhanced using hyperlinks.

Linked Data is an API when all is said an done. Its basically how we 
move from Open Database Connectivity (ODBC; an old platform specific API 
scoped to the extensional RDBMS dimension) to Open Data Connectivity 
(platform and data format agnostic variant of ODBC that also addresses 
the Intensional data dimension).

Most end-users and developers people seek enhanced Open Data 
Connectivity that leverage infrastructure and expanse of the World Wide 
Web.
>
> For the time being, http://graph.facebook.com/117527568273199
> illustrates my point (note that its also in JSON), and I want that to
> *count* as Linked Data.
Yes! It is Linked Data. It might not be 5-star Linked Data, but that 
doesn't matter right now. Its a powerful foundation that can evolve 
organically without even depending on Facebook.
> Right now that URI identifies *both* metadata
> about the movie and the movie itself. If/when Facebook wishes to mint
> a separate URI for the "metadata about the movie" from 'the movie
> itself" they should be able to, and then use w3c:describes OR 303
> redirection to connect the two URIs.

We implemented Linked Data middleware as part of our product portfolio 
because we understand there are many routes to Linked Data and its 
varying levels of fidelity. It's also why I continue to discourage 
conflation laden memes such as: RDF == Linked Data. Any structured data 
published to the Web is better than zilch plus endless arguments that 
span years.

Linked Data moniker problem:

When folks refer to 'Linked Data' it's important that we have some 
common understanding of what the moniker denotes. TimBL's meme picked a 
very generic term and that's lead to problems as the LOD Cloud and 
related efforts have exploded. I suggested to @danbri  -- in a different 
LOD list thread -- that maybe we use the term *hyperdata* to denote 
high-fidelity 5-Star Linked Data, for instance. This approach will 
further kill-off the distractions associated with 303 indirection 
debates re.,  hashless Linked Data URIs.

>
> As for the fact that might break OWL reasoners and semantics, I don't
> really care -  as one of the people who has done owl:sameAs reasoning
> on a large scale, you basically have to re-code everything in special
> purpose code to make things scalable (regardless of all the hype over
> decidable reasoning), so "OWL semantics" is kinda meaningless for
> practical purposes and as we have also shown, owl:sameAs is used in
> very odd contexts beyond strict identity all over the place.

Data integration is a practical purpose. Big Data is a moniker driving a 
meme that ultimately boils down to dealing with interwoven and 
exponential effects of the following re., data driven decision making:

1. data volume;
2. data velocity;
3. data variety (heterogeneity)
4. data location disparity.

Equivalence fidelity matters when dealing with data integration. This 
whole game is ultimately about data access, integration, and management 
cycles.

OWL (basic of deep) doesn't have to be at the front door. Folks don't 
have to publish high-fidelity Linked Data from the get-go. That said, we 
shouldn't knowingly encourage people to do the wrong thing either. An 
engaged end-user is an edge-case nightmare (in waiting) for developers. 
I've never bought into the 'dumb user' concept since I see them as being 
temporarily inattentive, until opportunity costs become palpable.

>
> I'm happy to talk philosophy (and have several papers about this) in
> an academic context, but from an engineering point of view I'd just
> like Linked Data to work.

It does work, and can work. We just need to be clear about the fact that 
it does have fidelity. We can also craft better narratives and 
exploitation road-maps.

IMHO. if 303 is at the front door of a Linked Data narrative, the 
narrative itself is broken. For instance, I don't see 303 at the front 
door of any version of TimBL's Linked Data meme.

> While Linked Data cloud is kinda huge, its
> not nearly as big as it *could* be if we did two things 1) replaced
> RDF/XML with a JSON format and 2) gave people easier options than 303
> redirections. I think those are the two major deployment barriers.

A Linked Data narrative that speaks about formats is equally broken :-)

The only things that matter re. high-fidelity Linked Data are the following:

1. entities have unambiguous names
2. names resolve to description (descriptor) documents
3. documents bear 3-tuples or triples based content that's constrained 
by the EAV (or RDF) data model
4. every document has a network address.

>
>     cheers,
>           harry


Links:

1. http://goo.gl/1dO33 -- about moving from Open Database Connectivity 
(ODBC) to Open Data Connectivity via Linked Data .

2. http://goo.gl/jhDcV -- using Google Spreadsheet to exploit Linked 
Data (basically an ODBC showcase demo reworked using Google Spreadsheet 
and SPARQL URLs).

Kingsley

>
>
>> //Ed
>>
>


-- 

Regards,

Kingsley Idehen	
Founder&  CEO
OpenLink Software
Company Web: http://www.openlinksw.com
Personal Weblog: http://www.openlinksw.com/blog/~kidehen
Twitter/Identi.ca handle: @kidehen
Google+ Profile: https://plus.google.com/112399767740508618350/about
LinkedIn Profile: http://www.linkedin.com/in/kidehen