Re: "undefined" URI scheme

On 09/11/2018 10:29, Martin J. Dürst wrote:
> Hello Laura,
>
> On 2018/11/09 19:07, Laura Morales wrote:
>> Hi, I'd like to suggest the creation of a new "undefined" URI scheme. It would be useful for any property that has to be defined as a URI, such as with RDF, but such property doesn't fit in any existing scheme nor has any well defined meaning associated with it like for example that of an RDF ontology. Roughly, the meaning would be that there exists such property but which meaning/scheme/context is not defined, so the user using it can interpret it however he likes, possibly by guessing the meaning.
>>
>> Example: <undefined:sensor-1>
>>            <undefined:length/vh>
>>
>> The closest thing to this that I could find is the <tag:taggingEntity:specific> scheme but the problem is that with this scheme the "taggingEntity" is mandatory, I cannot use for example <tag:sensor-1>. So I guess making the "taggingEntity" optional could also work instead of defining a new scheme, although <undefined:> would be more clear in my opinion.
>>
>> Thanks and sorry I've never contributed to this list before so I hope my request makes sense.
>
> I can understand where this comes from. But I think the dangers are big
> that more than one person would use <undefined:sensor-1>, with different
> meanings. That could wreck havoc with RDF. I think such a proposal
> should first be discussed in the RDF community, and brought here later
> when it's clearer what the exact semantics in RDF are.

+1

 From an RDF perspective, I'd suggest using an existing scheme; e.g.

     http://example.org/mystuff#sensor-1

which can at least dereference to a page that tells you that what it denotes is 
undefined and what expectations one might have of it if it were defined.  And 
maybe, at some future point, it could acquire a suitable definition (this being 
entirely consistent with RDF's "open world" approach to information modeling).

Note that in the semantics associated with RDF [1], the "meaning" (denotation) 
of a URI is not formally constrained by the URI scheme.

Another approach might be to use RDF blank nodes (which serve roughly as 
existential variables), but they are discouraged in some quarters, and not an 
option for properties [2].

Or use a UUID URI.

<aside>
I was initially going to suggest "http://example.org/undefined/sensor-1", but 
that could be misleading if the value later does acquire a definition.
</aside>

#g
--

[1] http://www.w3.org/TR/rdf11-mt/

[2] http://www.w3.org/TR/rdf11-mt/#blank-nodes

Received on Friday, 9 November 2018 13:45:20 UTC