Bernard,
I'm more than happy for Turtle to have a place in the LD ecosystem. My
concern is the suggestion that it should become seen as the primary/the
only delivery format for Linked Data resources such as www.lingvoj.org.
Also, like you, I'm not particularly impressed by massive dumps of
low-quality data. In my area of interest (cultural history) what we
need are reliable sources of data which is sufficiently richly
structured to be useful. This usually means they need to be more
structurally complex than a simple list of properties for an entity (the
"dbpedia fallacy").
One issue that Turtle will need to address (it may do so already) is
software support for free-hand data entry. While the format is
seductively simple-looking (well, it is to the likes of us who grew up
on XML/SGML*) it is very easy to make mistakes.
I followed Kingsley's reference to his file space (see his separate
reply) and grabbed the file jordan.ttl. It contains variations in
spelling which will mean that some predicate - subject links will fail
(e.g. New England Patriots), as will one sameAs link (USA). There is (I
guess) an intended link from USA to N. America, but again this won't fly
because USA's continent property is expressed as a string. If case
matters, most of the sameAs references won't work. The properties
(predicates) are all local to the document and none of them is defined.
Integer values are typed as strings. Two of the dates are wrong (e.g.
Sept 31 783). This is not to criticise Kingsley's typing, but rather to
point out that if you are encouraging users to hand-type resources which
are to be interpreted as data, then they are going to need some software
support if they are not going to be mightily let down by the whole
process. It's a bit like authoring web pages: it doesn't go too badly
if you're working in a rich edit box and don't have to add HTML markup
yourself.
Richard
* the Turtle tyro might disagree, on being asked to type e.g.:
:Championships "3 (2001,2003,2004)"^^xsd:string ;
On 06/02/2013 13:30, Bernard Vatant wrote:
Thanks Kingsley!
Was about to answer but you beat me at it :)
But Richard, could you elaborate on this view that hand-written and
machine-processible data would not fit together?
I don't feel like "people are still writing far too many Linked Data
examples and resources by hand". On the opposite seems to me we have
seen so far too many linked data produced by (more or less dumb or
smart) programs, without their human "productors" (so to speak) always
checking too much for quality in the process, provided they can
proudly announce that they have produced so many billions of triples
... so many, actually, that nobody will ever be able to assess their
quality whatsoever :)
Of course migrating automagically heaps of legacy data and making them
available as linked data is great, but as Kingsley puts it, linked
data are not only about machines talking to machines, it's also about
enabling people to talk to machines as simply as possible, and the
other way round. That's where Turtle fits.
Bernard
2013/2/6 Kingsley Idehen <[email protected]
<mailto:[email protected]>>
On 2/6/13 6:45 AM, Richard Light wrote:
On 06/02/2013 10:59, Bernard Vatant wrote:
More ??? Well, I was heading the other way round actually for
sake of simplicity. As said before I've used RDF/XML for years
despite all criticisms, and was happy with it (the devil you
know etc). What I understand of the current trend is that to
ease RDF and linked data adoption we should promote now this
simple, both human-readable and machine-friendly publication
syntax (Turtle). And having tried it for a while, I now begin to
be convinced enough as to adopt it in publication - thanks to
continuing promotion by Kingsley among others :)
And now you tell me I should still bother to provide n other
formats, RDF/XML and more. I thought I was about to simplify my
life, you tell me I have to make the simple things, *plus* the
more complex ones as before. Hmm.
Well I for one would make a plea to keep RDF/XML in the
portfolio. Turtle is only machine-processible if you happen to
have a Turtle parser in your tool box.
I'm quite happily processing Linked Data resources as XML, using
only XSLT and a forwarder which adds Accept headers to an HTTP
request. It thereby allows me to grab and work with LD content
(including SPARQL query results) using the standard XSLT
document() function.
In a web development context, JSON would probably come second for
me as a practical proposition, in that it ties in nicely with
widely-supported javascript utilities.
To me, Turtle is symptomatic of a world in which people are still
writing far too many Linked Data examples and resources by hand,
and want something that is easier to hand-write than RDF/XML. I
don't really see how that fits in with the promotion of the idea
of machine-processible web-based data.
Richard
--
*Richard Light*
If people can't express data by hand we are on a futile mission.
The era of over bearing applications placing artificial barriers
between users and their data is over. Just as the same applies to
overbearing schemas and database management systems.
This isn't about technology for programmers. Its about technology
for everyone. Just as everyone is able to write on a piece of
paper today, as a mechanism for expressing and sharing data,
information, and knowledge.
It is absolutely mandatory that folks be able to express triple
based statements (propositions) by hand. This is the key to making
Linked Data and the broader Semantic Web vision a natural reality.
We have to remember that content negotiation (implicit or
explicit) is a part of this whole deal.
Vapour was built at a time when RDF/XML was the default format of
choice. That's no longer the case, but it doesn't mean RDF/XML is
dead either, its just means its no longer the default. As I've
said many times, RDF/XML is the worst and best thing that ever
happened to the Semantic Web vision. Sadly, the worst aspect has
dominated the terrain for years and created artificial inertia by
way of concept obfuscation.
If your consumer prefers data in RDF/XML format then it can do one
of the following:
1. Locally transform the Turtle to RDF/XML -- assuming this is all
you can de-reference from a given URI
2. Transform the Turtle to RDF/XML via a transformation service
(these exist and they are RESTful) -- if your user agent can't
perform the transformation.
The subtleties of Linked Data are best understood via Turtle.
--
Regards,
Kingsley Idehen
Founder & CEO
OpenLink Software
Company Web:http://www.openlinksw.com
Personal Weblog:http://www.openlinksw.com/blog/~kidehen
<http://www.openlinksw.com/blog/%7Ekidehen>
Twitter/Identi.ca handle: @kidehen
Google+ Profile:https://plus.google.com/112399767740508618350/about
LinkedIn Profile:http://www.linkedin.com/in/kidehen
--
*Bernard Vatant
*
Vocabularies & Data Engineering
Tel : + 33 (0)9 71 48 84 59
Skype : bernard.vatant
Blog : the wheel and the hub <http://blog.hubjects.com/>
--------------------------------------------------------
*Mondeca*****
3 cité Nollez 75018 Paris, France
www.mondeca.com <http://www.mondeca.com/>
Follow us on Twitter : @mondecanews <http://twitter.com/#%21/mondecanews>
----------------------------------------------------------
Meet us at Documation <http://www.documation.fr/> in Paris, March 20-21
--
*Richard Light*
