Re: ISSUE-234: Started wiki page for response

I have updated the definition of "validation":

https://github.com/w3c/data-shapes/commit/af5aedc5d3d0b669b916757423ee5f330a2c8800

This is another attempt to address the ongoing issue that we are 
defining SHACL too procedurally, and gave the impression that all 
validation results MUST be produced at all times. The main change is 
that validation is defined to be a *mapping* between some input and 
validation results. I had already updated the definitions of components 
such as sh:node and sh:not to use the term "conformance checking" 
instead of validation, and added prose to make it super-clear that the 
results produced by such "nested" checking do not end up in the report.

I have also started to replace the formulation "... a validation result 
MUST be produced..." with "... there is a validation result...". So far 
I have only updated sh:class, because I would like to get your feedback 
on whether this a formulation that is both readable and "mathematically" 
acceptable.

My overall goal remains to produce a spec for SHACL that doesn't cause 
the readers to reject it as some rather theoretical gibberish.

Comments welcome,
Holger


On 27/02/2017 16:20, Holger Knublauch wrote:
> I have started a (long) wiki page to prepare the WG response to 
> Peter's latest list of comments:
>
>     https://www.w3.org/2014/data-shapes/wiki/ISSUE-234
>
> IMHO most issues were quite easy to address, many did not require 
> changes at all, some were essentially a matter of taste, others were 
> already discussed. So although he had indeed found some more glitches 
> I very much disagree with Peter's assessment that substantial changes 
> are required. In fact I believe this commit addresses most of the 
> things that he reported:
>
> https://github.com/w3c/data-shapes/commit/28f40d46efe714ebe9f1909f82e3fd84172dc447 
>
>
> I would like to discuss this on Wednesday and would appreciate input 
> on topics 36 and 37. Also, I would appreciate a review of my changes 
> to the textual definitions of sh:not, sh:and, sh:or, sh:xone, sh:node 
> and sh:qualifiedValueShape, which have been switched to use 
> "conformance" instead of "validation results". This helps avoid some 
> of the complications related to producing nested values as it makes it 
> clearer that implementers are not forced to produce nested values at all.
>
> Thanks,
> Holger
>

Received on Wednesday, 1 March 2017 08:22:27 UTC