Re: PROV-ISSUE-471 (wrong-wasAttributedTo-constraints): wasAttributedTo constraints not sensical [prov-dm-constraints]

Hi all,

It would be good to hear people's view on wasAttributedTo.  Should the agent
necessarily exist before the  entity was generated?

Should we also disallow examples such as:
   wasAttributedTo(painting,Bob,[prov:type="ownership"])

prov-dm is not precise about this, and we need to formalize some of 
these constraints in in prov-constraints.


I don't see how we cat prevent types of attribution such as this example,
without entering in a theory of causality (as Stian suggests).
   wasAttributedTo(painting,Bob,[prov:type="ownership"])

therefore, I think this example is legal prov-dm.


Luc



On 03/09/12 16:33, Stian Soiland-Reyes wrote:
> On Mon, Sep 3, 2012 at 2:52 PM, Luc Moreau <l.moreau@ecs.soton.ac.uk 
> <mailto:l.moreau@ecs.soton.ac.uk>> wrote:
> >> I don't agree with that. First of all, why has the attribution need to
> >> have anything to do with the invalidation of an entity? If you
> >> contribute to an entity, all of that has to happen *before* the entity
> >> is generated. It does not matter what happens after that.
> > Why should this be *before* the entity exist?
> >
> > I believe one can use attribution as follows:
> >
>
> I disagree.
>
> The DM spec (my highlights):
>
> Attribution^ ◊ <http://www.w3.org/TR/prov-dm/#concept-attribution>  is 
> the ascribingof an entity to an agent.
>
> When an entity e is attributedto agent ag, entity e was generatedby 
> some unspecified activity that in turn was associatedto agent ag. 
> Thus, this relation is useful when the activity is not known, or 
> irrelevant.
>
> An attribution^ ◊ <http://www.w3.org/TR/prov-dm/#dfn-wasattributedto> 
>  relation, written wasAttributedTo(id; e, ag, attrs) in PROV-N, has:
>
>   * id: an /optional/ identifier for the relation;
>   * entity: an entity identifier (e);
>   * agent: the identifier (ag) of the agent whom the entity is
>     ascribed to, and therefore bears some responsibility for its
>     existence;
>   * attributes: an /optional/ set (attrs) of attribute-value pairs
>     representing additional information about this attribution.
>
> Although attribute (v) 
> <http://www.oed.com/view/Entry/12933?rskey=xWNM2z&result=3&isAdvanced=false#eid> 
> in the wider sense does cover ownership:
>
>
>       *a.* To assign, bestow, give, concede, yield /to/ any one, as
>       his right (property, title, authority, worship,
>       honour)./arch./ or /Obs.
>       (..)
>       /*3.* To ascribe /to/ as belonging or proper; to consider or
>       view as belonging or appropriate /to/.
>
> the DM highlights "ascribe", as in:
>
>
>       *6.* To ascribe, impute, or refer, as an effect /to/ the cause;
>       to reckon as a consequence of.
>       *7.*To ascribe/to/an author as his work.
>
>
> This narrower understanding of 'ascribe' and 'attribute' was what I 
> had understood we are using, as we have not talked about ownership as 
> a kind of attribution before. We have been talking about a kind of 
> "why" or "who" made something appear - a book was written by an 
> author, a car was manufactured by a factory, a law was passed by its 
> parliament. There are many other definitions on "ascribe" and 
> "attribute" that I likewise don't think cover our intention with 
> wasAttributedTo, like: /ascribe great importance to or To ascribe as a 
> quality or ‘attribute’ belonging./
>
> If I own an old and dangerous car, I am not responsible for why it 
> /exists/, the car manufacturer is. I might bear responsibility for why 
> it has not yet been /invalidated/ as it is not road worthy, but that 
> has to do with potential future actions, intentions and plans, and I 
> don't see how /wasAttibutedTo/ in PROV would be suitable for that.
>
> We have said that PROV is provenance about the past. Describing that 
> kind of ownership would to me simply be an attribute on the entity, 
> just like it's location, colour, road worthiness status, insurance 
> status, who has access to the car keys, etc. Ownership would not in my 
> mind imply an activity (the "owning" activity? "purchasing"?), just 
> like having the colour red does not imply a "being red" activity. This 
> is about entity vs activity, state vs. change.
>
> If you want to broaden the definition of /wasAttributedTo/ to cover 
> mere 'ownership' kind of attribution, I think we need to add clear 
> examples that show the value of this and guides the understanding of 
> PROV-DM, and possibly reconsider the implied activity. I don't 
> remember us discussing this at a WG level.
>
> -- 
> Stian Soiland-Reyes, myGrid team
> School of Computer Science
> The University of Manchester

-- 
Professor Luc Moreau
Electronics and Computer Science   tel:   +44 23 8059 4487
University of Southampton          fax:   +44 23 8059 2865
Southampton SO17 1BJ               email: l.moreau@ecs.soton.ac.uk
United Kingdom                     http://www.ecs.soton.ac.uk/~lavm