Re: Can Shapes always be Classes?

On 11/19/2014 19:44, Simon Steyskal wrote:
> Although I completly agree with your arguments regarding the 
> redundancy of introducting a completely new formalism instead of 
> sticking to the "old" establish ones, I see the major advantage of 
> e.g. ShEx in its easy to read syntax and similarity to RegExp.
>
> I'm working partly on industrial related projects with engineers which 
> are not that familiar with semantic technologies but are somewhat open 
> minded regarding the use of those technologies for their use cases. 
> After I've attended this years SEMANTiCS and the headsup I gave them 
> regarding RDF validation/"CW reasoning"/ShEx/etc. they(we) are really 
> looking forward to the results of this WG (primarily because of the 
> aforementioned reasons).
>
> I think, if we want to reach even those data scientist/engineers/.. we 
> might want to use a language/formalism like ShEx (maybe in addition?) 
> which would ease their work.

Being "easy" to read is always subjective, and while I am sure that some 
engineers would be happy to work with regular expressions, other user 
groups won't. Instead of anecdotal evidence, we would need harder facts 
and those are difficult to get. But alienating the semantic web folks by 
disregarding SPARQL doesn't sound like a good starting point to me. 
After all, who is going to advocate and implement this new technology?

You mention that ShEx could be an addition to other languages. I assume 
you mean ShExC here, as the triple-based notation wouldn't really be 
attractive to newcomers. I can see that ShExC could sit as syntactic 
sugar on top of an RDF Shapes vocabulary (e.g. SPIN templates). But 
introducing yet another language has its own downsides:

- We create another tower of babel that is harder to learn
- Tools claiming Shapes support would need to support all layers
- We would end up with a family of "profiles", hindering interoperability

I wasn't part of the RIF WG, but I believe what happened there was that 
too many different directions were followed - every member of the group 
insisted to have their own dialect and requirements somehow supported. 
Not only according to my own opinion, the end result was a failure, and 
RIF is nowhere near a take-off. We should avoid repeating the same 
mistakes and not create another monster spec.

Holger

Received on Wednesday, 19 November 2014 23:06:41 UTC