Re: Serializations, mappings, and equivalence

Hi James,

Good questions. Here's my current understanding - but it could be wrong:

* PROV-DM is the definitive definition of the model. The ASN is there to 
help define the data model and is provenance specific. I believe, the 
idea was that it would easier to define the model because one need only 
worry about provenance and not how to fit it into OWL or some other 
modeling language.

* PROV-O defines a way to represent the data model using semantic web 
technologies. The by product of this is the ability to serialize model 
using RDF syntaxes.

* A serialization to XML was intended to be something that was "nice" 
xml - not RDF/XML

* We have not decided on what it means to be interoperable (i.e. 
equivalent) between serializations. This is what our interoperability 
discussion is about. See http://www.w3.org/2011/prov/wiki/Interoperability

Paul



James Cheney wrote:
> Hi,
>
>> From the minutes and discussion yesterday, it seems that the
>> question I asked yesterday may not have been understood, and later
>> discussion also touched on the point I was trying to make, so I
>> wanted to try again:
>
> 1.  The PROV-DM document uses a novel (but Prolog-like) abstract
> syntax notation (ASN), which can be used to describe instances of
> PROV.
>
> The PROV-O OWL ontology talks about RDF and gives constraints and
> facts about RDF instances of PROV, and uses RDF/XML.
>
> However, now I'm not sure what an instance of PROV "is".  Have we
> agreed that the RDF form is the canonical version?  I believe the
> charter implies this.   If so, is the ASN to be viewed as yet another
> compact syntax for RDF, and if so, why not use an existing one?
> Furthermore, are two instances of PROV equivalent exactly when their
> RDF versions are equivalent (modulo OWL reasoning)?  Or do we impose
> further rules for equivalence?
>
> If we have two different languages, then there is a danger - I think
> it's clear to *us* how to hand-convert from one representation to
> another, but it may be far from clear to people learning the model,
> and be a source of further confusion.
>
> The charter says, in comments under the deliverables:
>
> "A by-product of this formalization is the mapping of the provenance
> interchange language to RDF graphs."
>
> That is what I feel is missing - PROV-DM and PROV-O mention this
> mapping/serialization, but I am concerned that the part of PROV-O
> that describes this mapping does not give me (or someone unfamiliar
> with the model) enough information to take an ASN instance and
> convert it "blindly" to an RDF instance.  This could be something
> that the formal semantics formalizes - even if it is obvious to us,
> formally specifying it could be helpful to others for whom the
> mapping may not be obvious.
>
> 2. Similarly, there is a deliverable D5 on XML serialization:
>
> D5: PIL XML Serialization (W3C Note). This document specifies an XML
> serialization of PIL.
>
> and the comments say:
>
> "A serialization to XML (D5) will help disseminate the language to
> communities beyond the Semantic Web community."
>
> I understood this as meaning a direct serialization to XML that would
> not require use of RDF in places where it's not already common,
> whereas the discussion yesterday seemed to indicate that people think
> it would be fine to serialize as RDF/XML - i.e., if we standardize on
> RDF/XML then D5 is a no-op.  Is this correct?  If so, it seems to
> defeat the purpose of D5, and makes me wonder why we have an
> intermediate ASN instead of just using RDF everywhere.
>
> --James

Received on Friday, 21 October 2011 16:07:40 UTC