Re: Default RDF serialization

This is data independence.  The logical model (RDF) and the physical model(s) (serializations) need to be independent.

On Aug 23, 2012, at 2:19 PM, "Reza B'Far (Oracle)" <reza.bfar@oracle.com<mailto:reza.bfar@oracle.com>>
 wrote:

While I hate to sound like a broken record, I think at this point there is enough input from various folks on the thread (Arthur, Steve, Kingsley, David, Ashok, etc.) that the key is to separate model from serialization.  If that separation is made, then we can address things like performance, specific serialization formats (JSON-LD, etc.), potentially add new serialization formats in the future, enforce consistency between serialization formats via the model, etc.

So, IMHO, the first step would be to create a separation between serialization and model.  Otherwise, I don't see us coming to a consensus around which serialization format(s) should be selected (unless we select all of them in which case we'll have a few years worth of work to do).

Best.

On 8/23/12 12:13 PM, Arthur Keen wrote:

Can the LDP-WG consider performance as one of the criteria for selecting a default serialization?  RDF topology can factor into serialization performance.
For example, it can be very inefficient to serialize dense tabular data and time series data (measurements) into these RDF serializations.

Is there a way for us to have  triple-oriented serialization for sparse topologies and a tabular serialization for tabular RDF data in the same serialization?

Arthur


On Aug 23, 2012, at 10:31 AM, Kingsley Idehen <kidehen@openlinksw.com><mailto:kidehen@openlinksw.com>
 wrote:



On 8/23/12 10:19 AM, Steve K Speicher wrote:


I strongly agree as well with these points.  The only reason RDF/XML was
the only required serialization the member submission is it was the only
W3C Recommendation and we attempted to only reference "official"
standards.


Yes, I understand. Just as (after all these years) I'll never understand why the W3C hasn't acted on this most distracting and negative reality.

If RDF/XML's status as the sole syntax can't be addressed by putting Turtle on the same standing, then we have even more reasons for an overt and explicit loose coupling of RDF and Linked Data. Sadly, we have the complete opposite.



 Since this appears to be changing within W3C, then this
limitation no longer exists.  There was no technical reason, the
preference would be Turtle.  There is some consideration in the amount of
broad support for the serializations and therefore RDF/XML had a little
appeal but that is perhaps taking a too narrow view without the desire to
move things in right direction.


RDF/XML provides no benefits to folks that aren't building transformers. Folks like us build data (typically XML sources) transformers (cartridges) using RDF/XML, all of that happens behind the scenes and has no real impact on end-users and developers bar offering a plethora of formats for Linked Data Document content, via content negotiation.



I think you may have underestimated the problem you identified in (d) as
this is something that I deal with on a fairly regular basis.

My preference order for RDF serialization formats would be:
1) Turtle  (minimal requirement)


This is for end-users, integrators, and programmers.



2) JSON-LD


For JSON programmers.


3) RDF/XML


For XML programmers that understand RDF != XML.

Kingsley



Thanks,
Steve Speicher
IBM Rational Software
OSLC - Lifecycle integration inspired by the web ->
http://open-services.net<http://open-services.net/>

David Booth <david@dbooth.org><mailto:david@dbooth.org> wrote on 08/23/2012 10:08:05 AM:



From: David Booth <david@dbooth.org><mailto:david@dbooth.org>
To: public-ldp-wg@w3.org<mailto:public-ldp-wg@w3.org>,
Date: 08/23/2012 10:10 AM
Subject: Default RDF serialization

FWIW, if the LD profile is going to recommend one RDF serialization as
the default for RDF, I would argue strongly that it should be Turtle
instead of RDF/XML, because:

 (a) Turtle is far more human friendly to read;
 (b) RDF/XML is not XML Schema friendly;
 (c) RDF/XML has XML-based restrictions (such as prohibiting local names
that start with a digit) that make certain RDF difficult to represent;
 (d) RDF/XML has had a history of misleading developers who are familiar
with XML (but not RDF) into thinking that RDF is just a kind of XML.

Thanks!

--
David Booth, Ph.D.
http://dbooth.org/

Opinions expressed herein are those of the author and do not necessarily
reflect those of his employer.









--

Regards,

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