Re: RDF 1.1: "Some properties may change over time." (ISSUE-178)

On Tue, Dec 10, 2013 at 11:29:00PM -0600, Pat Hayes wrote:
> > Works for me. We could perhaps make it even simpler by just saying
> > 
> >   A relationship that holds between two resources at one time may not hold
> >   at another time.

Pat's answer (below) is certainly the more interesting.  However, the simple
bulleted list at [1] is not a good place to first raise such a subtle issue.
If Pat's judgement amounts to weak assent, I'd vote +0.5 for the "least bad"
variant above.

Tom

[1] https://dvcs.w3.org/hg/rdf/raw-file/default/rdf-concepts/index.html#change-over-time

> This last one is the least bad of the lot. But none of them are correct.
> There is a basic issue here. Just like sets, relations cannot really change
> with time. At least, not when they are described using a normal logic (they
> can in a tense logic). What can happen is that something that we might
> casually or carelessly describe as a binary relation is in fact a three-way
> relation with time as its third argument. Now of course [ R(a, b) at T ] or
> R(a, b, T) pretty much mean the same thing and in English we don't even have
> a way to distinguish them; but being all logical and strict about it, the
> three-argument way of talking is more accurate precisely because it makes it
> clear that the *actual relation* does not change, which makes sense because
> relations (speaking now formally and mathematically proper), like sets, just
> aren't the kind of thing that can possibly change. (If this reminds y'all of
> the problems we had with talking about RDF graphs being updated or modified,
> yes it is exactly the same issue.) We could have made RDF into a tensed
> logic, in which all assertions are made at a time, and things like a triple
> being true AT a time would make literal sense; but we didn't. So right now,
> and probably for the forseeable future, the idea of a relation changing -
> holding at one time but not at another time - does not make sense according
> to the RDF conceptual model, so temporal variation like this has to be
> modeled in the same way we would model a three-place relation in RDF. 
> 
> We might say something like this:
> 
> Some relations have an extra time parameter or are time-dependent. Such a
> relationship that holds between two resources at one time might not hold at
> another time. To describe this in RDF we have to treat the time as an extra
> argument or parameter to the binary relation. 

-- 
Tom Baker <tom@tombaker.org>

Received on Wednesday, 11 December 2013 06:18:41 UTC