Re: Another try.

On Feb 22, 2012, at 4:28 AM, Andy Seaborne wrote:

> ...
> I think the reality is that all existing RDF data is assuming "+" semantics.
> 
> Just putting a graph upon the web is an offer for people to decide whether to use it or not, and in what scope.  From the data consumer point of view, they use it under their acceptability criterion.


--------

On Feb 22, 2012, at 9:01 AM, Sandro Hawke wrote:

> On Tue, 2012-02-21 at 15:43 -0600, Pat Hayes wrote:
>> the rdf: and rdfs: vocabularies are not intended to be used
>> 'contextually' (eg they dont change with time and should be resistant
>> to subjective re-interpretation)
> 
> Is that really true?
> 
>    :SanFrancisco rdf:type :CityInCalifornia.
>    :CityInCalifornia rdfs:subClassOf :AmericanCity.
> 
> That seems like a fairly typical subclass relations, but it only became
> true in 1850.   There are current examples, too, of course.   Is that an
> error in modeling?  If so, I suspect it's a very common one.
> 
> The problem I'm having with this current discussion (which I think is
> headed in the right direction, to a point) is that I can think of
> reasons why pretty much *any* RDF triple might reasonably change over
> time.  By default, despite the RDF semantics, nearly every graph on the
> Web probably is in a different time-context, ....

---------

On Feb 22, 2012, at 9:47 AM, Dan Brickley wrote:

> On 22 February 2012 16:01, Sandro Hawke <sandro@w3.org> wrote:
>> On Tue, 2012-02-21 at 15:43 -0600, Pat Hayes wrote:
>>>  the rdf: and rdfs: vocabularies are not intended to be used
>>> 'contextually' (eg they dont change with time and should be resistant
>>> to subjective re-interpretation)
>> 
>> Is that really true?
> 
> Depends whose intention you're considering. I think the general
> consensus around W3C, from the start of this project, was that RDF/S
> *would* be used to describe things whose characteristics changed over
> time. So filesizes, city population counts, ages, even gender. ....

---------

Gentlemen. 

First, it was abundantly clear from the very beginning of the RDF WG activity that RDF/S (and DAML/OIL and subsequently OWL) were understood to be timeless logical languages. One can describe time-varying things in a timeless language, of course: you use times to do that, in fact. (You say "it is raining *at 3pm on the 14 March 2011*" instead of "it is raining" or "it is raining now".) And to do this does not require anything special in the underlying semantics, which is why RDF (etc.) don't have anything especially temporal built into their semantics. (Not because this would have been too complicated for the poor semanticians, who have been using temporal logics in KR work since around 1965.) What it does require, however, is either an extra argument (the time) to be added to relations - not a possibility for RDF 2004 - or else a rather complicated approach to event description. However, that approach is by no means impossible, and indeed is being used right now by several groups in one form or another (see [1] for an example, and a nice exposition.) But just adding a time parameter to relations is much easier and more natural, and doesnt require people to learn to think in a new way. And all it needs is allowing RDF to have 3-place relations as well as 2-place relations: not a very big step or indeed a very controversial one outside the RDF world. 

However, any way, simple or complicated, of describing time-varying stuff in a timeless logic does require users to be explicit about the time-relativeness of the data. If you are going to describe things whose characteristics change over time, then you do have to put the times in there somewhere (or else deal with the confusion and just plain factual errors that are going to arise if you don't). Because RDF is, and I suspect will always be, a timeless language. Adding an extra argument to relations does not change this basic fact about the language, it just makes it simpler to add that time parameter in what I think is a fairly natural way. So if Dan or anyone else is thinking that RDF is going to eventually become a time-embedded language in which assertions have a temporal state, then forget it. It isn't, because there arent any such languages; and so people who use it this way will have systems that break, and frankly, my dear, I don't give a damn. 

Second, regarding Sandro's point, while it might be true in some cosmic sense that nothing is permanent, the issue is whether we want to record data which is changing in a timesacle that matters relative to the use of our data. Im sure that there is going to be data that is useful and relevant now but will turn out, in hindsight, to be archaic a few centuries from now. But that does not matter if I am aiming at its being used a year from now rather than a century from now. 

BTW, changes in fact are not changes in meaning. Even if SF is not a city, the meaning of rdf:type does not change. If anything, what changed when SF got incorporated was SF itself. 

Finally, being time-relative is not the same as being uncertain or liable to be corrected if wrong, and being timeless is not the same as being ineffably correct. Of course data might have mistakes or even opinions embedded in it, and it might come from unreliable sources and so on. One might later correct errors in data that arose earlier. You might change your mind about something. You might weigh one source of information against another and decide to choose one and reject the other. But all of that has nothing to do with what the data is *claiming*. Imagine for the moment that we have some data which we know is absolutely reliable and correct and beyond reproach. In fact, we know it is *true*. Still the question arises, is it to be understood timelessly or as being time-embedded? Does it say that something is true "now" (in some suitably elastic sense) or is it a timeless assertion of a timeless fact? The distinction is still there even when our knowledge is perfect. 

Pat

[1] http://delos.zoo.ox.ac.uk/pub/2012/publications/Peroni-Shotton-&-Vitali_toc_eswc2012.pdf


------------------------------------------------------------
IHMC                                     (850)434 8903 or (650)494 3973   
40 South Alcaniz St.           (850)202 4416   office
Pensacola                            (850)202 4440   fax
FL 32502                              (850)291 0667   mobile
phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes

Received on Tuesday, 28 February 2012 15:55:08 UTC