W3C home > Mailing lists > Public > public-prov-wg@w3.org > September 2011

Re: Issue 89 - why?

From: Graham Klyne <GK@ninebynine.org>
Date: Wed, 21 Sep 2011 16:19:22 +0100
Message-ID: <4E7A007A.90402@ninebynine.org>
To: "Myers, Jim" <MYERSJ4@rpi.edu>
CC: Satya Sahoo <satya.sahoo@case.edu>, W3C provenance WG <public-prov-wg@w3.org>


On 19/09/2011 14:54, Myers, Jim wrote:
>>
>> Certainly - my point was that it doesn't prevent one from describing (say)
>> d1v1, d1v2, etc. and also separately saying that they arr "versions" of d1.
>> And it's a *lot* simpler than the current proposal.
>
> How so (re: it's simpler)?

Jim,

I think we may be discussing divergent topics.  My particular comment here was 
saying that OPMV is a lot simpler than the framework we're discussing by virtue 
of stating provenance only about static things.  The provenance model then just 
doesn't have be be concerned with what dynamic resource the annotated static 
content may be derived from - changes per se just don't enter the provenance 
annotation.

(I'm not specifically arguing for this, just using OPMV as an illustration that 
one doesn't need to know about fixed and variable attributes to make useful 
provenance assertions.)

#g
--

If you already have a class for document and a class for version, you just use 
the complementOf link between them - same as asserting a version relation. If 
you don't have classes (you don't control the domain ontology), or you don't 
want the provenance engine and users of provenance to have to know the temporal 
behavior of your classes, you can fix attributes/define a class from a base 
class and attribute constraints to expose them in a generic fashion at (just to) 
the level of precision needed to unambiguously interpret your provenance assertions.
>
> That's the only 'complexity' we're adding. Version gets used to talk about big changes to text, small changes, changes to the language used, changes in location, etc.  And we talk about a document with "Four score and seven years ago" in it after defining documents as mutable. Etc.  How can we answer simple questions about whether you have read those words if we don't know which things have fixed content? What mechanism other than a way to relate things that are mutable to immutable/less mutable versions of themselves (complementof) and a way to assert the values of the immutable properties of those things will work? (Just having the first part (complement/versionOf) and leaving the second to domain ontologists works in some sense - you can record provenance just as soon as those ontologies are complete.)
>
> All that said, I'm all for trying to explain the model in a more incremental fashion, e.g. starting from an OPM-like core, considering versions (as complementOf), and then introducing the idea of fixed attributes on entities as a way to generically document how the things in complementOf relationships differ and using that as a way to explain how to query, etc. (e.g. Asking for a document that contains a string has to be interpreted as finding a document who's latest version has that string or for which any version had that string. PROV is designed to make you pick one of the choices so you get an unambiguous answer ).
>
>   Jim
>
>
Received on Wednesday, 21 September 2011 15:51:17 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 26 April 2012 13:06:41 GMT