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

Re: PROV-ISSUE-183 (prov-dm-identifiers): identifiers in prov-dm [prov-dm]

From: Luc Moreau <L.Moreau@ecs.soton.ac.uk>
Date: Tue, 06 Dec 2011 17:31:56 +0000
Message-ID: <EMEW3|2139ad6da82502cc293d0abc3ff7b0d8nB5HW208L.Moreau|ecs.soton.ac.uk|4EDE518C.2080002@ecs.soton.ac.uk>
To: public-prov-wg@w3.org
Hi Simon,


On 12/06/2011 05:22 PM, Simon Miles wrote:
> Hi Luc,
> Isn't the consequence of your argument not only that two entity
> records in different accounts may have the same identifier, but also
> that two entity records in the same account may have the same
> identifier, as every entity record about one entity has the same
> identifier?
>    

That's not correct Simon.

There is a constraint that says that within an account, there is at most one
entity record for a given identifier
http://dvcs.w3.org/hg/prov/raw-file/default/model/ProvenanceModel.html#identified-entity-in-account.


> Your reasoning sounds to me more like a good argument against using
> the identifier for dual purposes than an argument in favour of scoped
> identifiers...
> > From an asserter's perspective, I like the suggestion from Paul's head
> that we reduce the burden of minting URIs simply by making record
> identifiers optional (but not scoped). But from a curator's
> perspective, I can see that it could be difficult to have provenance
> assertions we can't refer to or annotate later. For something more
> temporary than provenance, this might be fine, but I'm not sure it's a
> good idea here.
> I think overall, I prefer just to have the (light?) burden of minting URIs.
>    

If you go for minting URIs for entity records, then I don't think you
allow Paul to write his lightweight provenance as described in his blog.
I think it's a limitation.

Our original motivation for introducing qualified identifiers, which we
could restore if it is desirable, is that we had a construct, in 
prov-dm, to unambiguously
identify records (by combining their name with their account).

Luc

Luc

> Thanks,Simon
> On 6 December 2011 17:19, Luc Moreau<L.Moreau@ecs.soton.ac.uk>  wrote:
>    
>> Hi Paul,
>>
>> So, OK, we could mint identifiers for entity record
>>
>>            entity(<a minted identifier here>, [ex:param="a", ex:port="foo"])
>>
>> (Which by the way is what OPM does.)
>>
>> How do you refer to the entity now? We don't know what this record is about.
>>
>> Luc
>>
>> On 12/06/2011 05:11 PM, Paul Groth wrote:
>>      
>>> So I always thought that you could mint identifiers for entity records
>>> but you didn't have to and we supported that.
>>>
>>> But maybe that's my head inserting text where it shouldn't have been....
>>>
>>> Paul
>>>
>>> Luc Moreau wrote:
>>>        
>>>> ... the conclusion issue ;-)
>>>>
>>>> No, we have no formal decision on this.
>>>>
>>>> We wrote this in the prov-dm document a long time ago (before fpwd), and
>>>> we have
>>>> been refining it over time.
>>>>
>>>> I think it's an inevitable consequence of two key decisions:
>>>> - distinguishing entities (in the world) from entity records (in the
>>>> provenance)
>>>> - not mandating the minting of new URIs for entity records
>>>>        (no formal decision on this, but I think we have support for
>>>> it, since
>>>>         we want to minimize the effort to generate provenance)
>>>>
>>>> Luc
>>>>
>>>>
>>>> On 12/06/2011 04:56 PM, Paul Groth wrote:
>>>>          
>>>>> Hi Luc,
>>>>>
>>>>> Do you have a pointer to wear we reached the consensus about the dual
>>>>> role of identifiers?
>>>>>
>>>>> Thanks,
>>>>> Paul
>>>>>
>>>>> Provenance Working Group Issue Tracker wrote:
>>>>>            
>>>>>> PROV-ISSUE-183 (prov-dm-identifiers): identifiers in prov-dm
>>>>>> [prov-dm]
>>>>>>
>>>>>> http://www.w3.org/2011/prov/track/issues/183
>>>>>>
>>>>>> Raised by: Luc Moreau On product: prov-dm
>>>>>>
>>>>>>
>>>>>> Hi,
>>>>>>
>>>>>> It think that it is now time to have a proper debate about
>>>>>> identifiers in prov-dm since comments are regularly expressed about
>>>>>> them. I have raised this issue about this topic so that we can track
>>>>>> the conversation properly. Our hope is to reach consensus on this
>>>>>> topic by the time of the third working draft.
>>>>>>
>>>>>> First, in the fpwd, there was a mention of "qualified identifier"
>>>>>> (appearing in a note see [1]).  We have removed this term from the
>>>>>> second working draft.
>>>>>>
>>>>>> Second, the complementarity record now explicitly allows for linking
>>>>>> entity records across accounts. Its syntax allows for two accounts to
>>>>>> be named.
>>>>>>
>>>>>> Third, identifiers for entities in prov-dm have a dual role [3]. An
>>>>>> entity has got an id (typically given by an application). An entity
>>>>>> record --- i.e. what we say about an entity in a provenance record
>>>>>> --- also has an id. There is a consensus that we shouldn't mint
>>>>>> identifiers for provenance records. Hence, the identifier of the
>>>>>> entity record is defined to be the same as the identifier of the
>>>>>> entity.
>>>>>>
>>>>>> The consequence of this is that two entity records in different
>>>>>> accounts may have the same identifier: they may say different things
>>>>>> about the same entity.  For example, the document ex:doc was
>>>>>> generated by latex in account1, while in account 2, ex:doc is
>>>>>> described to be the result of a survey of a field by different
>>>>>> authors.
>>>>>>
>>>>>> This explains why we needed the complementarity record to name the
>>>>>> accounts as well. This assumes that account names need to be named
>>>>>> uniquely (see [4]).
>>>>>>
>>>>>> So, entity records identifiers are scoped to accounts.  Note, I said
>>>>>> entity *records*, not entities. Hence, we are not breaking the
>>>>>> semantic web approach: an entity is a resource and is denoted by a
>>>>>> URI, and this remains true in all accounts. (I guess that from a
>>>>>> semantic web perspective we are not looking at a provenance record as
>>>>>> resource, since we don't have a global URI to name it.) Finally, we
>>>>>> allow for accounts to be nested hierarchically; this fits nicely with
>>>>>> abstraction in provenance records. Again, see [4].
>>>>>>
>>>>>> Can you express your views about this approach, as currently defined
>>>>>> in the second draft of prov-dm?
>>>>>>
>>>>>> Thanks, Luc
>>>>>>
>>>>>> [1]
>>>>>> http://www.w3.org/TR/2011/WD-prov-dm-20111018/#expression-identifier
>>>>>> [2]
>>>>>> http://dvcs.w3.org/hg/prov/raw-file/default/model/ProvenanceModel.html#record-complement-of
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>              
>>>>> [3]
>>>>> http://dvcs.w3.org/hg/prov/raw-file/default/model/ProvenanceModel.html#record-Entity
>>>>>
>>>>>
>>>>>            
>>>>>> [4]
>>>>>> http://dvcs.w3.org/hg/prov/raw-file/default/model/ProvenanceModel.html#record-Account
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>              
>>>>          
>>>        
>> --
>> Professor Luc Moreau
>> Electronics and Computer Science   tel:   +44 23 8059 4487
>> University of Southampton          fax:   +44 23 8059 2865
>> Southampton SO17 1BJ               email: l.moreau@ecs.soton.ac.uk
>> United Kingdom                     http://www.ecs.soton.ac.uk/~lavm
>>
>>
>>      
>
>
>    

-- 
Professor Luc Moreau
Electronics and Computer Science   tel:   +44 23 8059 4487
University of Southampton          fax:   +44 23 8059 2865
Southampton SO17 1BJ               email: l.moreau@ecs.soton.ac.uk
United Kingdom                     http://www.ecs.soton.ac.uk/~lavm
Received on Tuesday, 6 December 2011 17:32:37 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 16:51:04 UTC