Re: Potential Formal Object from DERI over JSON-LD

* Richard Cyganiak <richard@cyganiak.de> [2012-10-22 19:17+0100]
> On 22 Oct 2012, at 18:32, Steve Harris wrote:
> >> Linked Data is not a protocol or format. It's a design pattern or an architectural style or a paradigm or whatever you call these abstract things.
> > 
> > Not according to http://www.w3.org/standards/semanticweb/data "What is Linked Data?"
> 
> Well, they might be selling something.

Indeed, and I think we'd all benefit from that "trademark" having a consistent meaning. Further, it probably has greater value to us if it is limited to things which have an explicit interpretation as an RDF graph.

Originally, RDF was a graph of URI References, Blank Nodes and Literals. Use of RDF implied the obvious good practices of sharing schemas, linking data by sharing entity identifiers, and backing those identifiers with GETtable resources. There was no label for these good practices.

Next, the RDF folks (perhaps TimBL?) coined the term "Semantic Web", underscoring the connectedness of RDF graphs. For a couple years, while it was principally associated with RDF, XML folks tried to apply it to pretty much anything with namespaces. That has settled down in the last few years and we can pretty confidently use the term "Semantic Web" to describe systems using the family of RDF technologies.

Most recently, TimBL coined the term "Linked Open Data" to categorize RDF data which was freely-available and reproducable and followed all of the RDF good practice. Removing "Open" leaves us with, well, RDF good practice.

I periodically complain within W3C about the lack of clear definitions for these terms. Fuzzy boundaries and inconsistent use makes it hard to utter them without a paragraph of explanation. Here is the most practical slicing of the definitions that I've come up with:
  3 Linked Data: Semantic Web data where identifiers (entity identifiers in particular) are GETtable.
  2 Semantic Web: RDF graphs sharing schemas and entity identifiers.
  1 RDF: graph of IRIs, Blank Nodes and Literals.

This specifically moves "Linked Data" away from the slippery slope of its more general interpretation as any data which shares some identifiers (which could be argued to include HTML). I honestly don't see any place else to draw the line and I don't know what good the term does the community if it doesn't carry any consistent interpretation.


> > You view Linked Data as a paradigm, I view it as a set of interoperable standards.
> 
> You're prepared to categorically exclude certain technologies from the set of standards that constitutes Linked Data. I'm not.
> 
> The “set” is constantly evolving, and has consistently included technologies that were not (yet, sometimes) W3C standards. If W3C working groups only get involved after the fact, then who gets to say what's in the set and what is not? TimBL? You? Me?
> 
> Like the Web, Linked Data is a system in which many different formats, protocols and vocabularies can coexist. Interoperability comes from clients that make choices about which subset of the standards to support.
> 
> RDF is the big fish in the Linked Data pond, just like HTML is the big fish in the WWW pond. This doesn't mean that other formats are not part of the Linked Data web or are not part of the WWW. Having tried both, I can say with confidence that explaining why RDF is a good format for Linked Data works *much* better than insisting that other formats are not Linked Data.
> 
> Best,
> Richard
> 
> 
> 
> > 
> > - Steve
> > 
> >> FYI, LDP-WG is defining a protocol with minimum requirements for doing Linked Data with RDF.
> >> 
> >> On 22 Oct 2012, at 16:27, Antoine Zimmermann wrote:
> >>> Now, I sympathise with Kingsley leitmotiv that we (individual persons, not WG) should present RDF in ways that suit our audience's vocabulary (e.g., EAV rather than SPO). But this has to be done in tutorials, classrooms, presentations, communications, etc with our partners, not inside the W3C spec, IMO.
> >> 
> >> +1
> >> 
> >> Best,
> >> Richard
> >> 
> >> 
> >> 
> >>> 
> >>> 
> >>> -AZ
> >>> 
> >>> 
> >>> Le 22/10/2012 17:05, Steve Harris a écrit :
> >>>> For what it's worth (and this is really off topic, especially for the
> >>>> RDF working group) I don't agree with this viewpoint at all.
> >>>> 
> >>>> IMHO if we don't define a lingua franca for Linked Data then we're
> >>>> wasting our time even discussing the idea of it. The situation is bad
> >>>> enough with RDFa, JSON-LD, GRDDL, POWDER, Ntriples, N3, Turtle, and
> >>>> RDF/XML - not to mention NQuads, TriG, and TriX - without bringing
> >>>> other data models into the equation as well.
> >>>> 
> >>>> http://www.w3.org/standards/semanticweb/data makes no mention of
> >>>> non-RDF stack languages.
> >>>> 
> >>>> - Steve
> >>>> 
> >>>> On 2012-10-22, at 14:03, Richard Cyganiak wrote:
> >>>> 
> >>>>> Hi Peter,
> >>>>> 
> >>>>> If it can be sent over HTTP, and can express hyperlinks in a
> >>>>> standard way, and can express arbitrary attribute-value pairs in a
> >>>>> standard way, then it can do Linked Data.
> >>>>> 
> >>>>> Best, Richard
> >>>>> 
> >>>>> 
> >>>>> On 22 Oct 2012, at 12:03, Peter F. Patel-Schneider wrote:
> >>>>>> So I could have reverse Polish notation
> >>>>>> http://en.wikipedia.org/wiki/Reverse_Polish_notation linked data?
> >>>>>> Or object-oriented http://en.wikipedia.org/wiki/Object_database
> >>>>>> linked data? Or fuzzy http://en.wikipedia.org/wiki/Fuzzy_logic
> >>>>>> linked data?  Or Montague
> >>>>>> http://en.wikipedia.org/wiki/Montague_grammar linked data?  These
> >>>>>> are all standard in some sense.
> >>>>>> 
> >>>>>> There has to be some notion that everyone is serving up stuff
> >>>>>> that others can read.  Otherwise linked data is nothing more than
> >>>>>> a slogan.  But where is the boundary?  It seems to me that the
> >>>>>> boundary is triples, i.e., RDF.  What linked data adds is nothing
> >>>>>> more than pragmatics.  (Not that pragmatics isn't important.)
> >>>>>> 
> >>>>>> peter
> >>>>>> 
> >>>>>> 
> >>>>>> On 10/22/2012 06:02 AM, Richard Cyganiak wrote:
> >>>>>>> Pat,
> >>>>>>> 
> >>>>>>> On 22 Oct 2012, at 04:59, Pat Hayes wrote:
> >>>>>>>> I would be very interested to discover what y'all consider
> >>>>>>>> the be the definition of Linked Data. Can you provide a
> >>>>>>>> pointer to where this can be found? Thanks in advance.
> >>>>>>> http://en.wikipedia.org/wiki/Linked_Data
> >>>>>>> 
> >>>>>>> Trying to nail it down much tighter than that is
> >>>>>>> counter-productive. I learned this the hard way a couple of
> >>>>>>> years ago, when foolishly trying to stop people who were “doing
> >>>>>>> Linked Data with Atom” from using the LD term.
> >>>>>>> 
> >>>>>>> One can obviously do Linked Data with RDF, and that's by far
> >>>>>>> the most popular approach. RDF is well-suited to that task, and
> >>>>>>> it's the community where the LD term first emerged. W3C's
> >>>>>>> LDP-WG is currently writing a specification that has more
> >>>>>>> details for that.
> >>>>>>> 
> >>>>>>> JSON-LD is an attempt at creating a format that allows doing
> >>>>>>> Linked Data with JSON. That's great. However, there's a thin
> >>>>>>> line between saying “we enable LD with JSON” and “JSON-LD is
> >>>>>>> how you do LD”. The JSON-LD spec really ought to say only the
> >>>>>>> first thing, but slips into implying the second too often.
> >>>>>>> 
> >>>>>>> Attempting to enforce a particular implementation technology
> >>>>>>> for Linked Data, be it RDF or JSON or Atom or Microdata or
> >>>>>>> whatever, doesn't work. This is what Kingsley keeps repeating
> >>>>>>> on a daily basis, and he's right.
> >>>>>>> 
> >>>>>>> The fact that a JSON-LD document also can be parsed to an RDF
> >>>>>>> graph is mostly orthogonal to this.
> >>>>>>> 
> >>>>>>> Best, Richard
> >>>>>> 
> >>>>>> 
> >>>>> 
> >>>>> 
> >>>> 
> >>> 
> >>> -- 
> >>> Antoine Zimmermann
> >>> ISCOD / LSTI - Institut Henri Fayol
> >>> École Nationale Supérieure des Mines de Saint-Étienne
> >>> 158 cours Fauriel
> >>> 42023 Saint-Étienne Cedex 2
> >>> France
> >>> Tél:+33(0)4 77 42 66 03
> >>> Fax:+33(0)4 77 42 66 66
> >>> http://zimmer.aprilfoolsreview.com/
> >>> 
> >> 
> >> 
> > 
> > -- 
> > Steve Harris, CTO
> > Garlik, a part of Experian
> > +44 7854 417 874  http://www.garlik.com/
> > Registered in England and Wales 653331 VAT # 887 1335 93
> > 80 Victoria Street, London, SW1E 5JL
> > 
> > 
> 
> 

-- 
-ericP

Received on Tuesday, 23 October 2012 12:05:58 UTC