W3C home > Mailing lists > Public > public-prov-wg@w3.org > February 2012

Re: PROV-ISSUE-239 (TLebo): is wasGeneratedBy(e) meaningful? [prov-dm]

From: Stian Soiland-Reyes <soiland-reyes@cs.manchester.ac.uk>
Date: Tue, 7 Feb 2012 11:17:49 +0000
Message-ID: <CAPRnXt=MkJThD66Ag7A4TnshBAkCobM5Skrvx=WyVtN9k-0Yzg@mail.gmail.com>
To: Daniel Garijo <dgarijo@delicias.dia.fi.upm.es>
Cc: Paolo Missier <Paolo.Missier@ncl.ac.uk>, Timothy Lebo <lebot@rpi.edu>, "public-prov-wg@w3.org" <public-prov-wg@w3.org>
On Tue, Feb 7, 2012 at 10:12, Daniel Garijo
<dgarijo@delicias.dia.fi.upm.es> wrote:

> what if instead of the time, I just knew the location (it was generated in
> Madrid),  or the role of the activity (i.e., creator), but not the activity itself?

(you mean role of the entity, right)


> Maybe we are running into a corner case, but as Stian suggested I don't see
> any reason why wasGeneratedBy([],e) should be forbidden.

Agree, we should not rule out partial provenance just because it does
not fit our ideal ASN syntax.


> Now that I realize, we are not supporting this in prov-o!

This has been mentioned before, PROV-O support it through bnode hoops:

:entity prov:wasGeneratedBy [
    a prov:Activity ; # which we know nothing more about
    prov:hadQualifiedGeneration [
      prov:hadLocation <http://dbpedia.org/resource/Madrid> ;
      prov:hadRole :someRole
    ]
] .


which I'm kind-of OK with. I admit I added prov:wasGeneratedAt to
PROV-O to avoid this big structure for merely specifying the
generation time, but now you were quite specific and should be
prepared to use a qualified involvement.

wasGeneratedAt seems a route we are thinking of going down in PROV-DM.
It does not work for usage time, though. You want to know when an
entity was used in a given role, with a given activity. You could
state simply that an entity was used, without knowing the activity,
but I don't think we need new syntax for that corner case.


The semantics of PROV is that an entity (within an account/bundle/xx)
can only be generated once, by one activity. So it does not matter if
we know about the activity or not, we can easily talk about it
indirectly, and even infer that any later assertions about the
generation activity is the same one.

Similarly you can talk about an activity being controlled by (or
associated with) an agent, even if you don't know the agent. For
instance you might be on a tram and assume there is an agent driving
it (prov:role :Driver), but it could even be an autonomous software
agent if you didn't go to the front of the tram to check.


-- 
Stian Soiland-Reyes, myGrid team
School of Computer Science
The University of Manchester
Received on Tuesday, 7 February 2012 11:44:42 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 26 April 2012 13:06:56 GMT