W3C home > Mailing lists > Public > public-ldp@w3.org > December 2012

Re: Links and graphs

From: Kingsley Idehen <kidehen@openlinksw.com>
Date: Tue, 18 Dec 2012 07:51:26 -0500
Message-ID: <50D066CE.2040704@openlinksw.com>
To: public-ldp@w3.org
On 12/17/12 5:54 PM, Erik Wilde wrote:
> hello kingsley.
> On 2012-12-17 10:13 , Kingsley Idehen wrote:
>> I believe LDP is based on the following:
>> 1. RDF data model -- an entity relationship data model endowed with
>> explicit machine and human comprehensible semantics
>> 2. Linked Data -- structured data based on the RDF model
>> 3. HTTP -- data access protocol that's decoupled from data
>> representation formats
>> 4. RDF data formats -- Turtle (MUST) and others (MAYBE).
>> If the above is true, why do you keep on recycling the same debate
>> topics in different guises?
> i am simply trying to keep us from making avoidable mistakes. HTTP is 
> not a data access protocol. 

And this an example of where there is profound confusion. Of course HTTP 
is a data access protocol. If not, then what is it?

> if it were, we would just use FTP. 

FTP is a different data access protocol.

> the "H" is there for a reason: HTTP clients are supposed to navigate 
> through an interlinked set of resources, which make their interaction 
> capabilities discoverable through HTTP's uniform interface. as soon as 
> we start treating HTTP as data access, and address issues in RDF that 
> should be addressed on the HTTP layer, we make ourselves incompatible 
> with those 99% of today's web who don't speak RDF. 

HTTP is a Data Access Protocol, and Linked Data is all about 
accentuating that very point. Basically, every URI is a Data Source 
Name. Every URL is a Data Source Address/Location.  Data is delivered in 
a variety of negotiable formats.

HTTP based Linked Data is an improvement over RDBMS specific Data Access 
Protocols such as ODBC, JDBC, ADO.NET, OLE-DB etc..

The above holds true even if you aren't working with RDF data (resources).
> if we "hide" issues necessary for interaction in RDF instead of 
> exposing it in HTTP concepts, we make ourselves invisible to the 
> majority of users that we see as possible adopters of LDP.

You continue to infer that RDF is just about data representation and 
serialization formats. This is because (I think) you are overlooking the 
fact that it's builds on the well established entity relationship data 
model courtesy of the following:

1. use of HTTP URIs to denote Entities
2. use of triple patterns to express Entity Relationships
3. use of triple patterns to express fine-grained Entity Relationship 

Note: RDF based Linked Data just adds the de-reference requirement to 
the URIs referred to in point #1.

There is nothing to hide re., use of RESTful patterns (for Create, Read, 
Update, and Delete) to deductively interact with  data spaces capable of 
processing RDF payloads on an network that supports HTTP. Net effect, 
smart clients, servers, and peers orchestrated using RESTful patterns. 
Application sophistication can rise without exponential increases in 
conventional programming language specific code. Basically, its 
ultimately about more Data and less Code, since the data is 
self-describing and said descriptions ultimately amounts to logic that 
drives computing.

Data is finally becoming the Program.

> afaict, we would treat short-term convenience of not trying to be good 
> HTTP citizens, for the long term disadvantage that we would be just 
> another well-behaving web service (see mark baker's earlier comments 
> about "shouldn't we try to be friendly to anybody speaking HTTP?", and 
> of course we should).

Nothing about RDF makes it incompatible with the Web as it exists. We 
only get into trouble when we use RDF as the moniker for everything 
thereby leading to massive confusion at every turn.

I haven't seen a single comment in these threads -- from those that 
support RDF -- that's incompatible with good HTTP practice and broad 
compatibility. Even mentions of SPARQL remained baked in URIs since this 
very sophisticated intensional query language fits naturally into HTTP 
re. actual query language constructs, data access service, and query 
results serialization formats.

BTW -- some of us have used AtomPub for years to move RDF payloads 
between clients and servers over HTTP, it hasn't caused us any problems 
either :-)

> cheers,
> dret.



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

Received on Tuesday, 18 December 2012 12:51:53 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:03:09 UTC