Re: PROV-ISSUE-8: defining generation in terms of `IVPT of'

Luc Moreau wrote:
> Graham,
>
> Can you suggest an alternate definition?

The gist of my intervention was to question that we need any such definition, so
I'm not sure that I'm morally bound to suggest an alternative :^)

It doesn't help that I've lost track of what definition(s) we were discussing:
I was responding in particular to Khalid's suggestion "we add a property to
IVPT...".  IPVT is a term we've agreed to use pro tem, but once we start adding
properties to this term, I fear it becomes locked in to the outcome of our
discourse without actually serving any useful purpose.

That said, I think we're talking about "derivation" and/or "generation".

I would adopt "derivation" as a relationship between resources (e.g. R2 is a
derivation of R1 -- without explicit reference to IVPT, but understanding
implicitly that a resource might be an IVPT of some other resource), according
to which is is claimed that R1 has made a significant contribution to the nature
of R2.  I've stolen a leaf from James' book here to explicitly allow that the
exact nature of claim is subjective, or not completely defined;  it is
sufficient that the claim is admitted into the discourse concerning the
relationship between R1 and R2.

Much more productive, I think, than defining the exact nature of this
relationship is to explore the kinds of properties we expect such a relationship
to exhibit, that coincide with our intuitive or subjective view of the
relationship.  E.g. if (R2 derivedFrom R1) and (R3 derivedFrom R2) is there
anything useful we can claim about the relationship between R1 and R3?  (I think
this is the sort of approach that is taken with OPM, which seems quite sensible
to me at this level.)

I think "generation" would be similar, but as a relation between a process
execution and a resource.

...

I'm reminded by a recent TAG message about a presentation Pat Hayes gave to a 
WWWC workshop in Edinburgh a few years ago...

'There's a hint in the first paragraph of the section entitled "IF
AMBIGUITY IS INEVITABLE, LET'S MAKE LEMONADE"'
-- http://lists.w3.org/Archives/Public/www-tag/2011Jun/0056.html

 From the abstract of the referenced paper:
"Reference is by nature ambiguous in any language. So any attempts by Web 
architecture to make reference completely unambiguous will fail on the Web. 
Despite popular belief otherwise, making further ontological distinctions often 
leads to more ambiguity, not less."
-- http://www.ibiblio.org/hhalpin/homepage/publications/indefenseofambiguity.html

While I recognize the problem addressed here is somewhat different in character 
to the ones we are addressing, I think that the lessons of attempting to 
over-constrain the meaning of a term still have some relevance.  So I am 
suggesting that at some level we have to allow a a term in outr definitions to 
stand with merely intuitive, or subjective, interpretation.

(Model theoretic semantics leads clearly to a similar conclusion:  a "merely" 
formal definition can never completely constrain the interpretation of the terms 
used in an expression (a wff of some language) - what they can do is provide a 
formal basis for expressing, and reasoning about, the properties of statements 
using those terms.)

#g
--


Luc Moreau wrote:
> Graham,
> 
> Can you suggest an alternate definition?
> 
> Professor Luc Moreau
> Electronics and Computer Science
> University of Southampton 
> Southampton SO17 1BJ
> United Kingdom
> 
> On 10 Jun 2011, at 18:19, "Graham Klyne" <GK@ninebynine.org> wrote:
> 
>> It's not clear to me yet that the model *needs* to distinguish these cases, even if we can recognize them.
>>
>> My quote of the day comes from the schema.org debate:
>> "In my experience, metadata design efforts tend to fall into the trap of focusing more about what could be said about a topic rather than what needs to be said in order to support use cases of the consuming software."
>> -- Henry Sivonen, http://hsivonen.iki.fi/schema-org-and-communities/
>>
>> #g
>> --
>>
>>
>> Khalid Belhajjame wrote:
>>>   Hi Jim and Luc,
>>> I agree with Luc, Jim point is a good one. I find it more relevant to derivation than generation, though. Generally, derivation can be though of as a relationship that connects an IVPT of a thing to another IVPT of the same or different thing. I can only think of two options to deal with the point raised by Jim. Either:
>>> - we add a property to IVPT that identify the thing that the IVPT gives a view about, or
>>> - specialize the derivation relationship, by creating two sub-relationships that distinguish between the two cases.
>>> Personally, I prefer the second one, as it spares us the problem of having to identify “thing”, at least for the moment.
>>> Thanks, khalid
>>> On 10/06/2011 08:09, Luc Moreau wrote:
>>>> Hi Jim,
>>>>
>>>> *very* good questions, that's the essence of IVPT, I think.
>>>>
>>>> I don't have answers, and need to think about this.
>>>>
>>>> I was looking at Generation alone, you seem to allude to Derivation.
>>>> Their definitions may need to be drafted together.  I will think about this.
>>>>
>>>> Luc
>>>>
>>>> On 10/06/11 02:28, Myers, Jim wrote:
>>>>> This would mean that a heating process modifies an egg to create a warm egg, it does not transform a cold egg into a warm egg?
>>>>>
>>>>> Or do you mean both - a process execution can turn one thing into another, these things can be considered IVPTs of a thing that participates in the process execution/ is modified by the process execution? And in an open world assumption, a witness doesn't have to report the modified thing or can decline to identify/report either of things in IVPT roles depending on their ability to observe and the use case they wish to enable?
>>>>>
>>>>> ________________________________
>>>>>
>>>>> From: public-prov-wg-request@w3.org <mailto:public-prov-wg-request@w3.org> on behalf of Luc Moreau
>>>>> Sent: Thu 6/9/2011 6:44 PM
>>>>> To: Provenance Working Group WG
>>>>> Subject: PROV-ISSUE-8: defining generation in terms of `IVPT of'
>>>>>
>>>>>
>>>>>
>>>>> Hi all,
>>>>>
>>>>> - if a new thing is created, it is clear that we have a new IVPT of that thing
>>>>>
>>>>> if a chicken creates an egg is it just an IVPT of an egg?
>>>>>
>>>>> - if the thing is modified, then it is a requirement that a new view (IVPT) is generated ...
>>>>>        otherwise, it would still be a view that existed before
>>>>>
>>>>> can't I say the egg was heated without reporting its cold and warm states? I.e. don't we want to be able to report that something was modified without having to report the IVPTs? A document was edited four times by different people but I don't wan't to/can't tell you what each wrote at each stage?
>>>>>
>>>>> - if the process execution was taking a long time to modify/create the thing, there is only one
>>>>>    instant at which the (invariant!) IVPT appears
>>>>>
>>>>> I thnk we could define it that way, but if a cracking process takes time, saying the cracked egg appears instantaneously basically means you want 'cracked egg' to be defined by some threshold - the cracked egg might become more cracked over time ) invariant only in that it is always above the threshold and the instance of the creation of the IVPT relationship occurs ata  aspecific instant.
>>>>>
>>>>> - I think this captures well a stateful objects, where processes can modify the object, resulting in
>>>>>    different IVPTs corresponding to the various states
>>>>>
>>>>> IVPTs are not a separate kind of thing and their invariance is relative. If they are truly immutable sates/snapshopts, they can only exist for an instant because some part of the state of the thing (a part we may not care about such as age) will change immediately.
>>>>>
>>>>> What do you think?
>>>>>
>>>>> http://www.w3.org/2011/prov/wiki/ConceptGeneration#Definition_of_Generation_by_Luc 
>>>>>
>>>>> Cheers,
>>>>> Luc
>>>>>
>>>>>   
>>>>
> 
>

Received on Saturday, 11 June 2011 22:49:08 UTC