Re: Request for feedback about RDF-star

On 25/03/2021 17:47, Kjetil Kjernsmo wrote:
> torsdag 25. mars 2021 15.43.10 CET skrev Pierre-Antoine Champin:
>>> My suggestion is to mint a URI outside of the rdf: namespace, but make
>>> it clear in any notes or WDs that it is expected that it will change.
>> As I hinted in my original email, we know from experience that this does
>> not work, unfortunately.
> 
> Yes, I agree, there's a lot of precedence, but the world is quite different
> from back then. The rdf: namespace was a very early one, the FOAF effort was
> a long-term evolving effort. This is around 20 years ago.
> 
> Now, we have learnt to wrap namespace maps into reusable modules that can
> be used throughout the codebase, so that a change needs to happen in just
> one place. Right...? Well, come to think of it, a transition from one
> namespace to another would still be more painful than just change its URI,
> but still.

The problem is not so much the code base handling the data, it is the 
data itself. Once you have loads of RDF resources on the Web using the 
"old" IRIs, deprecating that IRI essentially breaks these resources. And 
you can not expect everyone to update their content at once.

So in order to ensure backward compatibility, it is not enough for your 
code to switch from the old namespace to the new one. It must recognized 
the two namespaces as synonymous. And that is significant trouble for 
implementers.

> 
> BTW, I just realized that my SKOS example was a bad one, they actually
> changed back. :-)
> 
> I think the world is sufficiently different now for this to work out fine up to
> CR, 

That's assuming that the whole Web is updated on a regular basis, which 
it is not...

> but it is just an opinion, not a research-based advice.

Same for me ;)

> 
> Cheers,
> 
> Kjetil
> 
>