Re: Implementation feasibility

On 3/22/15 3:17 PM, Jose Emilio Labra Gayo wrote:
> It is better now to put every language construct on the table so the 
> WG can select which ones should or should not be part of the language. 
> That's why I to propose to start designing the SHACL high-level 
> language instead of imposing an implementation based on SPARQL.

I do not believe this group is capable of anticipating all future 
requirements that average SHACL users will need. A far better process is 
therefore to provide a generic mechanism that allows anyone to define 
new language constructs. The SHACL template mechanism is exactly that - 
any community can define new terms, back them with formal semantics and 
share their definitions in machine readable format.

There are some obvious template candidates such as sh:minCount and 
sh:valueType that will be included by default. But there is nothing 
substantially different between sh:minCount and, say, my:keyProperty, or 
(as suggested today on the Hydra mailing list) the ability to narrow 
down properties by namespace, or to state that all literals must have 
distinct language tags. There is an almost infinite variety of such 
constraint patterns, and SPARQL covers them nicely.

The design of a hard-coded language with fixed semantics for a random 
subset that some small group of people in 2015 found interesting is 
fundamentally outdated and does not exploit the spirit of the semantic web.

Holger

Received on Sunday, 22 March 2015 09:19:38 UTC