W3C home > Mailing lists > Public > semantic-web@w3.org > July 2014

Re: Should information be merged from several RDF files?

From: Eric Prud'hommeaux <eric@w3.org>
Date: Mon, 7 Jul 2014 17:07:56 -0400
To: Victor Porton <porton@narod.ru>
Cc: Pat Hayes <phayes@ihmc.us>, "semantic-web@w3.org" <semantic-web@w3.org>
Message-ID: <20140707210741.GC19116@w3.org>
* Victor Porton <porton@narod.ru> [2014-07-07 21:13+0300]
> Sorry, my example was wrong as a :tranformer may have more than one :script-data.
> 
> However a transformer can't have more than one precedence, and multiple precedences may arrive if we merge several RDF files.
> 
> So, anyway I consider stupid merging several RDF files wrong (or hard to implement properly).

I understand that you need to impose some rules to keep your application logic managable, but I hope you don't conclude that RDF itself discourages merging data or offers any notion of completeness of data. Innumerable use cases for RDF involve merging complementary data from different sources. Your rule is a proxy for issues around authority and correctness, and as such provides an interesting use case for RDF validation.


> 07.07.2014, 20:12, "Victor Porton" <porton@narod.ru>:
> > Finally now I think that I should consider incomplete data in an RDF file as erroneous (and ignore it), even when combined from several files it becomes complete. Below there is an example that, generally, combining several RDF files may lead to a nonsense.
> >
> > An example supporting this behavior:
> >
> > # In one RDF file
> > <http://example.org/example-transformation>
> >         a :transformer ;
> >         :source-namespace <http://example.org/ns1> ;
> >         :target-namespace <http://example.org/ns2> ;
> >         :precedence <#macro> ;
> >         :script-data [
> >                 a :xslt ;
> >                 :script-url <http://example.org/scripts/foo.xslt> ;
> >                 :transformer-kind :entire ;
> >                 completeness 0.9 ;
> >                 stability 0.9 ;
> >                 preference 0.9
> >         ] .
> >
> > # In the second RDF file
> > <http://example.org/example-transformation>
> >         a :transformer ;
> >         :source-namespace <http://example.org/ns1> ;
> >         :target-namespace <http://example.org/ns2> ;
> >         :precedence <#macro> ;
> >         :script-data [
> >                 a :xslt ;
> >                 :script-url <http://example.org/scripts/bar.xslt> ;
> >                 :transformer-kind :entire ;
> >                 completeness 0.9 ;
> >                 stability 0.9 ;
> >                 preference 0.9
> >         ] .
> >
> > This would make both foo.xslt and bar.xslt associated with http://example.org/example-transformation but this is a nonsense. With a transformation only one :script-data can be associated.
> >
> > So if we follow open world assumption, after downloading the first RDF it would be OK, but after downloading also the second RDF it would become an error. One thing it's bad for, is that this would complicate program logic (as we would need to remove an already filled data structure which contains info about http://example.org/example-transformation).
> >
> > To make my program logic non-contradictory, I probably should ignore later loaded data about http://example.org/example-transformation if it was already extracted from an RDF loaded earlier.
> >
> > It seems that open world assumption makes my program logic contradictory!
> >
> > 06.07.2014, 18:02, "Pat Hayes" <phayes@ihmc.us>:
> >>  On Jul 5, 2014, at 10:07 AM, Victor Porton <porton@narod.ru> wrote:
> >>>   I work on this project:
> >>>   http://freesoft.portonvictor.org/namespaces.xml
> >>>
> >>>   It involves downloading RDF files from the Web.
> >>>
> >>>   When a "bundle" (for lack of a better word), that is a set of related RDF triples
> >>  I really do not know what you mean by 'related' or 'bundle'. Neither of these terms make RDF sense.
> >>>   , is incomplete it should be ignored.
> >>  That is a very odd rule to use when dealing with RDF. It is a basic aspect of the RDF design that information from various sources can be combined, and that no single source should be assumed to be complete (the 'open world' assumption). So you are using RDF here in a way which, while not formally incorrect, does violate the design presumptions of RDF.
> >>>   My question, if a bundle is split into several different RDF files, and each part of it is incomplete, should it nevertheless not be ignored if the union of all parts of the bundle is complete?
> >>  I would say, clearly no. Which illustrates why this 'ignore' rule is troublesome.
> >>
> >>  Pat Hayes
> >>>   A toy example (A, B1, B2, C be namespaces):
> >>>
> >>>   In one file:
> >>>
> >>>   <A> <B1> <C> .
> >>>
> >>>   In an other file:
> >>>
> >>>   <A> <B2> <C> .
> >>>
> >>>   If both B1 and B2 properties are required, should this information be ignored (as incomplete bundles)? or should the information from the two files be merged and thus considered complete?
> >>>
> >>>   --
> >>>   Victor Porton - http://portonvictor.org
> >>  ------------------------------------------------------------
> >>  IHMC                                     (850)434 8903 home
> >>  40 South Alcaniz St.            (850)202 4416   office
> >>  Pensacola                            (850)202 4440   fax
> >>  FL 32502                              (850)291 0667   mobile (preferred)
> >>  phayes@ihmc.us       http://www.ihmc.us/users/phayes
> >
> > --
> > Victor Porton - http://portonvictor.org
> 
> --
> Victor Porton - http://portonvictor.org
> 

-- 
-ericP

office: +1.617.599.3509
mobile: +33.6.80.80.35.59

(eric@w3.org)
Feel free to forward this message to any list for any purpose other than
email address distribution.

There are subtle nuances encoded in font variation and clever layout
which can only be seen by printing this message on high-clay paper.
Received on Monday, 7 July 2014 21:08:02 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 1 March 2016 07:42:52 UTC