Re: PROV-ISSUE-1 (define-resource): Definition for concept 'Resource' [Provenance Terminology]

martin wrote:
> Hello Simon, Graham,
> 
> I regard indeed talking about immutable things as a good starting point. 
> If we take this, I have no problems.
> 
> However, the consequence is that Information Resources cannot undergo 
> change, otherwise they become undefined.
> In other terms, their state must be is fixed to their identity.
>  So, provenance of an information resources becomes a chain of 
> derivation, whereas provenance of a physical object
> can imply changes of state without change of identity, such as the 
> etching plates of Rembrandt and the Bartsch numbers
> for their states.
> 
> So, I am happy with "immutability" as upper definition, as long as we 
> are aware that some necessary properties
> for Provenance will differ significantly for material things, as Graham 
> points out.

My comment is that I was *less* concerned about immutability...

> A concept like "freedom" or "sparrow" may also be regarded as being not 
> immutable, due to the lack of
> identifiable substance, multiplicity of opinions about their substance 
> and fluctuations of meaning. A good reason to exclude concepts.

I think it's fair to make statements about the origin of concepts, but as you 
say there may be a multiplicity of opinions.  But the Web is like that anyway. 
That suggests a need provenance for assertions of provenance too.  Of course, 
some such statements may not be susceptible to the kinds of analysis envisaged 
by the formal provenance models we've seen to date.

> Talking about anything that can be identified I believe is insufficient, 
> because this implies things like
> the processes themselves we talk about.

Thinking about preservation of workflow-related information, that seems to me to 
be a reasonable goal.

#g
--

> On 5/31/2011 11:26 AM, Graham Klyne wrote:
>> I find the approach described by Simon is quite close to the views I'm 
>> trying to crystalize and articulate. I would probably be less
>> definitive about the immutability of resources.
>>
>> While it's sometimes useful to be able to talk about physical vs 
>> information objects, I think the boundary isn't always so clearly defined
>> in a way that works for all scenarios. Ontologies like CIDOC CRM have 
>> a view that works well in their domain. But I'd be wary of forcing the
>> CRM model onto a general provenance framework, even though a 
>> provenance framework must be able to accommodate the CRM view. At some 
>> point,
>> the provenance framework must work with, and must "hand over", to 
>> other domain ontologies - and making too much ontological commitment
>> within a provenance framework could make that less easy in some cases.
>>
>> So, my starting point would be that provenance in general can be about 
>> anything that can be identified. But then there are certain aspects
>> of provenance covered on both OPM and PML and others frameworks that 
>> focus on capturing relationships between particular kinds of resource:
>> when these relationships are described they imply constraints on the 
>> resources to which they apply, but which are not necessarily
>> constraints on all possible uses of provenance. I think such an 
>> approach would make it easier to accommodate and reason about different
>> models of provenance, their commonalities and differences. In the 
>> first instance, I would look for the most general interpretations.
>>
>> #g
>> -- 
>>
>> Simon Miles wrote:
>>> Hello Martin,
>>>
>>> I wouldn't disagree with what you say about physical and information
>>> resources, but I would take a different perspective on what needs
>>> defining.
>>>
>>> The view taken in OPM (and, before that, the PASOA project and
>>> others), is that to unambiguously talk about a "thing's" provenance,
>>> the "thing" you are talking about must be immutable. The provenance of
>>> a mutable thing would depend on its state when you ask the question,
>>> and this might not even be the state you think it is (it may have
>>> changed since you last looked). This is not to say that it is
>>> unreasonable for anyone to ask for the provenance of something
>>> mutable, only that this should be answered by drawing on the more
>>> tangible provenance of immutable things.
>>>
>>> Either a "physical object" or an "information object" can be viewed
>>> from a single state and context, i.e. its state unchanging at a given
>>> time and (for information objects) replica. This state is immutable
>>> (once the object changes it becomes a new state) and has a
>>> well-defined provenance.
>>>
>>> I agree that the provenance of the state of a physical object may look
>>> different from the provenance of a state of an information object, in
>>> that prior states of the same object will be linear for the physical
>>> object, but may not be for the information object. However, I'm not
>>> convinced that this is actually important. The provenance of anything
>>> will surely be non-linear, as what it is now will derive from not only
>>> prior versions of itself, but also whatever else caused it to change
>>> or be created (e.g. I am a product of my earlier self, but also of
>>> your emails :-)).
>>>
>>> So, my personal inclination would be to start with defining a concept
>>> for something immutable, and specify what the provenance of such a
>>> thing would be, then expressing what the provenance of something
>>> mutable is based on that.
>>>
>>> I would also personally not predefine what an immutable thing could
>>> be. For example, one webpage contained in one HTTP response may appear
>>> to be an immutable state, but could be transformed by proxies in
>>> communication or by the browser after download or have copies cached,
>>> while on the other hand, we may be so certain the batch of items from
>>> a production line are identical and unchanged through transit that we
>>> can unambiguously talk about them as one immutable state even though
>>> in spans space (multiple objects) and time (time in transit). Another
>>> example is that "reading" a file will not change its contents, but may
>>> change a timestamp which may later be relevant to how it is used.
>>> Immutability seems critical to defining provenance, but anything is
>>> immutable only by ignoring changes you consider irrelevant.
>>>
>>> Thanks,
>>> Simon
>>>
>>> On 29 May 2011 14:02, martin <martin@ics.forth.gr> wrote:
>>>> Dear All,
>>>>
>>>> Following the phone-conference on May 26, let me repeat some thoughts:
>>>>
>>>> The definition of a Resource that has the potential to have a 
>>>> provenance
>>>> (following Guarion, Gruber, ontologies describe possible states of 
>>>> affairs as precisely as possible)
>>>> in a Semantic Web relevant way, should be specific enough, so that 
>>>> we can clearly
>>>> identify a set of properties that are relevant and connect in a 
>>>> relevant way to
>>>> answer provenance query.
>>>>
>>>> From our background I suggest that the distinction of a Physical 
>>>> Resource consisting of matter
>>>> ( for instance crm:E18 Physical Thing) from an InformationResource 
>>>> (irw:InformationResource
>>>> or crm:E73 Information) is necessary and fundamental, because
>>>>
>>>> 1) a physical thing undergoes a linear sequence of states and 
>>>> changes, because any change destroys
>>>> the previous form. It can only be at on place at a time. From this 
>>>> we infer most of our
>>>> common sense logic of provenance and identity. Even splitting or 
>>>> merging an object destroys all
>>>> predecessors. Identity can be based on continuity of custody 
>>>> (sequence of all ID cards), or
>>>> essential properties (fingerprints etc).
>>>>
>>>> 2) An Information Object can reside on multiple carriers (or 
>>>> "realizations", "copies", "items") at the same time.
>>>> The state of change of any of the copies cannot be related without 
>>>> complete world knowledge to that of
>>>> other copies, because we cannot know what may happen on the other 
>>>> side of the world.
>>>> Therefore the Information Object itself has no well-defined or 
>>>> verifiable states of change in its nature as data.
>>>> Therefore changes of Information Objects are better described as 
>>>> creations of new ones for any minimal change.
>>>> Identity can be based on content, for provenance reasoning best on a 
>>>> bit or character identity.
>>>>
>>>> As a consequence, analogue photograhic material in film industry 
>>>> etc. is better traced as material objects,
>>>> because there is no convention to define identity of content for 
>>>> different copies of analogue photographs.
>>>>
>>>> Using provenance for authenticity reasoning on information objects 
>>>> will rely, besides others, on the fate of multiple copies.
>>>> Not being able to distinguish the behavior of carriers from the 
>>>> actual data would be prohibitive to
>>>> such reasoning.
>>>>
>>>> Further, universals ("Concept", crm:E55 Type), such as "man", "dog" 
>>>> behave again differently, because the
>>>> IsA relations and often fuzzy boundaries of concepts create again 
>>>> different identity conditions and much
>>>> more confused states. I propose to exclude provenance of universals 
>>>> from the discussion until we have understood
>>>> the other two.
>>>>
>>>> I maintain that no more distinctions need be made for this 
>>>> PROVENANCE discussion.
>>>>
>>>> FRBR entities have been mentioned in the discussion. In the CRM-FRBR 
>>>> Harmonization Group we
>>>> concluded 
>>>> (http://www.cidoc-crm.org/docs/frbr_oo/frbr_docs/FRBRoo_V1.0.1.doc)
>>>> together with the IFLA FRBR Review Group that the identification of 
>>>> Work, Expression, Manifestation
>>>> is in practice done by selecting "representative" existing 
>>>> realizations, which have a clear identity by content,
>>>> be it fragments or copies of copies of lost works. Therefore the 
>>>> "conceptual nature" of a Work should not confuse
>>>> us. The provenance would still be based on realizations.
>>>>
>>>> Best,
>>>>
>>>> Martin
>>>>
>>>>
>>>> -- 
>>>>
>>>> --------------------------------------------------------------
>>>> Dr. Martin Doerr | Vox:+30(2810)391625 |
>>>> Research Director | Fax:+30(2810)391638 |
>>>> | Email: martin@ics.forth.gr |
>>>> |
>>>> Center for Cultural Informatics |
>>>> Information Systems Laboratory |
>>>> Institute of Computer Science |
>>>> Foundation for Research and Technology - Hellas (FORTH) |
>>>> |
>>>> Vassilika Vouton,P.O.Box1385,GR71110 Heraklion,Crete,Greece |
>>>> |
>>>> Web-site: http://www.ics.forth.gr/isl |
>>>> --------------------------------------------------------------
>>>>
>>>>
>>>>
>>>> ______________________________________________________________________
>>>> This email has been scanned by the MessageLabs Email Security System.
>>>> For more information please visit http://www.messagelabs.com/email
>>>> ______________________________________________________________________
>>>>
>>>
>>>
>>>
>>
>>
>>
> 
>

Received on Tuesday, 31 May 2011 14:16:36 UTC