I can't manage the detailed context(sic) of this discussion and example when it 
comes round every few days -- but I do believe that the contextualization 
relation is potentially dangerous.

As far as I can tell it has NO semantics that distinguish it from 
specializationOf, yet I believe it encourages users to read into it uses that 
could violate the semantics of RDF URI usage.

Thus, my immediate question is this.  What inferences are possible using 

    contextualizationOf(e1, e2, b)

that are not also valid inferences using specializationOf

    specializationOf(e1, e2)


I can't see any.  In the absence of such, I can't see any valid reason for 
including contextualizationOf.  And in the absence of such, I can't see the 
formal basis for your (implied) claim that contextualizationOf does identify the 
bundle in which ex:Bob occurred.  I believe this is an area in which we really 
*need* formalization and rigour.

Part of the reason that I'm so wary of this particular relation is that I think 
it usurps a part of semantic web technology that is being defined by the RDF 
working group ("named graphs", datasets and associated semantics).  As such, I 
think the whole discussion about this should be conducted in the provenance+RDF 
coordination group.


On 21/06/2012 13:43, Luc Moreau wrote:
> Hi Graham,
> It is on purpose that we didn't define context. In the WG there is a precedent
> of not defining
> terms. To name a few: thing, aspect, act upon/with, bear responsibility, goals,
> steps, assignment, ...
> So the word 'context' is to be understood with its informal meaning, .... like
> 'context' already
> used in the definition of association [1].
> As far as your response to Tim is concerned, of course, we
> can say that if
> tool:Bob_as_in_run1 prov:contextualizationOf ex:Bob
> then
> tool:Bob_as_in_run1 prov:specializationOf ex:Bob
> given that contextualizationOf is defined as a subproperty of specializationOf.
> But your solution fails to identify the bundle in which ex:Bob occurred, which
> led to the poor rating.
> Regards,
> Luc
> [1]
> On 15/06/2012 16:32, Graham Klyne wrote:
>> This is my attempt to respond to ACTION-92.
>> On 14/06/2012 23:07, Luc Moreau wrote:
>> > In your absence, we assigned ACTION-92 to you.
>> > Can you provide an example of contextualization you think
>> > may break rdf semantics?
>> This is hard to do without a complete formal description of what
>> contextualization actually means.
>> I did reply [1] to Tim's comment, and the reference to the example in the wiki.
>> I would adjust my earlier comment to say something like this: I cannot see
>> how contextualization can be anything but vacuous without violating RDF
>> semantics; i.e. how it actually expresses anything that cannot be expressed
>> without it.
>> I've studied the description of contextualization in DM
>> (,
>> and when I dig in to it I find I can't make any sense of what it is saying.
>> As it stands, the notion of context is undefined, so I am unable to interpret
>> statements like "A bundle's descriptions provide a context in which to interpret
>> an entity in a domain-specific manner". What is this "context"? When I look to
>> the definition of "bundle", I see "A bundle is a named set of provenance
>> descriptions ...". There's nothing here about defining or providing a
>> "context". So this notion of context is being introduced without any grounding
>> or basis for understanding what it means.
>> #g
>> --
>> [1]
>> On 14/06/2012 23:07, Luc Moreau wrote:
>>> Graham,
>>> In your absence, we assigned ACTION-92 to you.
>>> Can you provide an example of contextualization you think
>>> may break rdf semantics?
>>> Thanks,
>>> Luc
>>> PS. Tracker, this is ISSUE-385
>>> On 14/06/12 23:04, Luc Moreau wrote:
>>>> Dear all,
>>>> Minutes of todays's teleconference can be found at
>>>> Thanks to Paolo for scribing.
>>>> Regards,
>>>> Luc

Received on Friday, 22 June 2012 15:31:27 UTC