Re: blank slate

Holger,

Yes, your example is most likely understandable by anyone who does some 
coding, without being a "semantic web engineer." Now the question is: 
who creates this and how? And do they have to be fully versed in RDF?

One of my goals for RDF "application profiles" (which to me include 
defining shapes or containers, as well as specifying validation) is that 
the structure itself should guide novices to make good RDF-compatible 
decisions. I see the result (at least of the Dublin Core work in this 
area) possibly as being templates to be followed, as well as a 
dictionary of reusable common modules. Some of those modules could be 
very simple (minimum cardinality) and others could be quite complex. 
Those who don't need the highly complex patterns can just ignore them. 
The original Dublin Core Description Set Profile constraint language was 
defined as an XML XSD, but it was designed as a set of templates that 
others could follow in developing their own application profile. [1]

I've seen RDF/XML and, to a lesser extent, JSON-LD act as gateways to 
RDF. (I almost wanted to say: gateway drugs.) The down side of that is 
that users tend not to learn to think in terms of key RDF concepts like 
graphs, inferencing, and the meaning of class membership in RDF. So it 
does help that there are known serializations, but those known 
serializations can hide the ways that RDF semantics differ from our 
previous ways of thinking about data. Therefore people tend to create 
perfectly good XML or JSON data that isn't RDF at heart.

I had thought that at least one of the goals of this group was to 
provide a standard way to define structures and constraints that will 
result in "good" RDF, and that the outcome was to serve as wide of a 
user group as possible. RDF suffers still from a lack of usable tools 
for non-RDF experts.

As for the experts on this list, I often wonder, since some seem quite 
content with either SPARQL or OWL or applications that build on these, 
what their need is around this standard. Since we do have a mix of 
communities represented here, maybe we should share our needs before we 
discuss solutions.

kc
[1] http://dublincore.org/documents/dc-dsp/



On 8/4/14, 5:25 PM, Holger Knublauch wrote:
> Do you think JSON-LD might be an acceptable compromise for your users,
> or not?
>
> Roughly like (untested):
>
> {
>      "@context" : "http://w3.org/2014/shapes/context",
>      "@id" : "ex:Person",
>      "constraint" : {
>          "type" : "shapes:Property",
>          "property" : "ex:firstName",
>          "occurs" : "shapes:Exactly-one",
>          "valueType" : "xsd:string"
>      }
> }
>
> Holger
>
>
> On 8/5/2014 10:06, Jeremy J Carroll wrote:
>> I share some of Karen's concerns.
>>
>> At Syapse we are integrating our RDF based systems with legacy systems
>> such as HL7 systems.
>> The engineers dealing with the legacy systems don't think RDF, don't
>> know RDF, they don't want to know RDF.
>> They don't think open world or closed world, etc.
>>
>> Given that the legacy system (like many legacy systems) is (overly?)
>> complicated we don't want to add a layer of over-complication of our
>> own; however we do want to have expressive power to deal with the
>> complications in the patterns if that is appropriate.
>>
>> I feel that this use case (HL7/RDF integration) shares some of the
>> issues that Karen seems concerned about: the pattern language is being
>> used to communicate with engineers who are not RDF literate, and do
>> not want to be.
>>
>> Jeremy
>>
>>
>>
>> On Aug 4, 2014, at 1:48 PM, Markus Lanthaler
>> <markus.lanthaler@gmx.net> wrote:
>>
>>> On 3 Aug 2014 at 17:19, Karen Coyle wrote:
>>>> Actually, the "compact, human readable syntax" is what I am most
>>>> interested in. It may need to be built on top of what the group
>>>> develops, but without it, the community I am most interested in will
>>>> not
>>>> be able to participate, as we will have few members with the technical
>>>> skills to express constraints in something resembling, for example, a
>>>> complex SPARQL query.
>>>>
>>>> I posted a reply to this thread that no one has replied to, so it is
>>>> sitting there sadly orphaned. Briefly, what I do not see anywhere in
>>>> this conversation any mention of WHO is the target of this
>>>> "deliverable".
>>> That's indeed a very important question that has, IMO as well, been
>>> mostly ignored so far.
>>>
>>>
>>>> There is a great deal of discussion of the technology but
>>>> almost none of the real world in which it will operate, and zero
>>>> discussion of the target skill set of the intended implementers. As so
>>>> often seems to happen in standards work, the skill set of the
>>>> members of
>>>> the standards group is assumed as the target skill set of all users.
>>> Since this group is working on RDF validation and not JSON or XML
>>> validation, I think it is fair to assume at least some knowledge of
>>> RDF. As such, I think a "compact, human readable RDF-based syntax" is
>>> a very reasonable thing. I'm not too much a fan of introducing yet
>>> another (serialization) format/syntax.
>>>
>>>
>>> --
>>> Markus Lanthaler
>>> @markuslanthaler
>>>
>>>
>>>
>>>
>>
>
>
>

-- 
Karen Coyle
kcoyle@kcoyle.net http://kcoyle.net
m: 1-510-435-8234
skype: kcoylenet

Received on Tuesday, 5 August 2014 01:24:04 UTC