Re: issue-95 metamodel simplifications

I find the current SHACL syntax to be overly complex.  I think that it will
turn out to be a significant barrier to SHACL use unless it can be hidden
behind tools or a compact syntax.

The complexity of the metamodel is mirroring the complexity of syntax.  In my
view a simpler syntax with a simpler metamodel will be easier for users, even
if it is more verbose than the current syntax in some cases.


On 03/02/2016 03:51 PM, Holger Knublauch wrote:
> (Meta-comment): I feel this discussion (see Peter's parallel emails on this
> topic) is going in the wrong direction. It seems that some people find the
> metamodel complex, and therefore consider changing the language itself so that
> the metamodel becomes less complex. But the metamodel is barely visible to
> anyone - only expert users who want to create their own extensions ever have
> to worry about it. OTOH the user-facing syntax of SHACL is what everyone will
> see, and it was designed driven by use cases and requirements. IMHO, if these
> use cases and requirements cause a certain complexity in the metamodel, so be
> it. (Honestly I also don't find the metamodel draft complex, but that's just
> my opinion).
> Eric, when you speak about "filters", are you referring to sh:filterShape or
> sh:constraint? If you are referring to sh:filterShape, where does this even
> show up in Proposal 3?
> Holger
> On 3/03/2016 9:24, Eric Prud'hommeaux wrote:
>> Looking through the metamodel proposals┬╣, it seems that a lot of the
>> complexity comes from the fact that triple constraints, inverse triple
>> constraints, and filter constraints are different beasts. What are the
>> use cases that motivate filters?
>> The XML Schema WG provided only one way, a processing instruction, to
>> associate XML elements with types in a schema. Years later, WSDL added
>> another, associating parts of the REST I/O for some endpoint with
>> types in some schema. The processing instruction seems analogous to
>> the sh:classShape property connecting instances of some class to a
>> given shape. The WSDL analog would fit naturally in some LDP doc.
>> ┬╣
>> <>

Received on Wednesday, 2 March 2016 23:56:16 UTC