Re: prov-dm attributes: two proposals to vote on (deadline Wednesday midnight GMT)

+1 on Proposal 2.

+0.5 on Proposal 1.

I fully support it if it doesn't break anything fundamental and I'm
pretty sure it doesn't, but just have small unnamed doubts.
Characterisation seems important from the perspective of two asserters
knowing they are asserting about the same thing, and for
distinguishing the different provenance of, for example, a document
from a version of the document. The practical consequence of the
proposal seems to be that all attributes could be considered
characterising by some actor, which sounds very plausible.

As I've argued before, I would prefer to go even further and say that
not just attributes but all *assertions* (including provenance
assertions) about an entity are potentially characterising. For
example, the assertion that DocV2 was generated by EditingPE1 is not
an attribute of DocV2, but does distinguish it from all other entities
(including Doc and DocV1). As a querier or different asserter, I can
use the wasGeneratedBy assertion to know whether the entity is what
I'm looking for/want to assert about, no differently to if I'd
searched for attribute-value pairs. There is no reason this argument
would apply to only provenance assertions like wasGeneratedBy and not
assertions about an entity more generally.

Thanks,
Simon

On 31 October 2011 00:56, Reza B'Far (Oracle) <reza.bfar@oracle.com> wrote:
> +1 on Proposal 1.
> No comment on Proposal 2.
>
> On 10/30/11 10:54 AM, Luc Moreau wrote:
>
>
>
> In the interest of simplification, we would like to make the following
> proposal about attributes in prov-dm.
>
> Proposal 1: attributes are a necessary part of prov-dm. Attribute-value
> pairs can be optionally
>   included in Entity Expressions and Activity Expressions.
>
> The document will justify their presence along the following lines
> (text to be worked on, suggestions welcome). For inter-operability
> purpose, it is necessary to be able to describe entities (and
> activities), and such descriptions need to be part of the provenance
> record, so that queries over such descriptions can be expressed.  The
> document will not make the distinction between non-characterizing and
> characterizing attributes. All attributes will be considered as
> describing some facet of the entity.
>
> Proposal 2: Constraints related to attributes will be dropped.
> http://www.w3.org/TR/2011/WD-prov-dm-20111018/#derivation-attributes
> http://www.w3.org/TR/2011/WD-prov-dm-20111018/#use-attributes
> http://www.w3.org/TR/2011/WD-prov-dm-20111018/#generation-affects-attributes
>
> Rationale: a number of issues have been raised against these
> constraints. They may or may be fixable. But overall, they seem to
> overconstraint the model, for benefits that are unclear.  There was no
> intent to make them automatically verifiable, for
> instance. Furthermore, if it is really crucial for some developers to
> express that some attributes depend on others, than prov-dm already
> offers a mechanism: simply model these attributes as entities, and their
> dependency as derivation.
>
> A further consequence is that derivation can be made transitive!
> (subject of a separate proposal)
>
> Finally, the semantics team may want to reconsider these constraints and
> formalize them properly.
>
> Cheers,
> Luc
>
>
>



-- 
Dr Simon Miles
Lecturer, Department of Informatics
Kings College London, WC2R 2LS, UK
+44 (0)20 7848 1166

Received on Monday, 31 October 2011 11:07:47 UTC