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

Re: PROV-O comments

From: Simon Miles <simon.miles@kcl.ac.uk>
Date: Fri, 2 Dec 2011 17:31:38 +0000
Message-ID: <CAKc1nHc8F2pUQ3FvLr6Q=M765NdKzUzNDh1v0oLnJ09UkAUagA@mail.gmail.com>
To: Provenance Working Group WG <public-prov-wg@w3.org>
Hi Stian,

I agree with all you say in your email, but think that the Prov-O
draft contradicts that.

Prov-O, Section 3.1.4:
  The Instant class represents "point-line" temporal information that
have "no interior points" [OWL-TIME]
Section 3.2.13.1, startedAt:
  This object property defines the time when Activity started. The
time is specified as an time:Instant [OWL-TIME].
(and similarly for endedAt and generatedAt).

so startedAt etc. require the asserter to refer to a time that has no
interior points, which she is unable to do, as it is not under her
control to enforce whether it has no interior points, for all the
reasons you give in your mail.

I don't think we need to "cater for the quantum physicist at all
granularities", and should concern ourselves with more likely problem
cases, e.g. where generation is asserted at the granularity of a day
because that is what the asserter knows, but later critical
differences are made between what was true at different points in the
day.

I think this is solved by just using something less restrictive than
Instant, so not implying anything unenforceable ("no interior
points").

Disagreement (as you refer to with your third asserter) is a separate
matter - any asserters could give mutually inconsistent accounts
regardless of time.

Thanks,
Simon

On 29 November 2011 23:35, Stian Soiland-Reyes
<soiland-reyes@cs.manchester.ac.uk> wrote:
> I think philosophically and physically you can never truly have a
> non-divisive "instant" unless it is defined at Planck time. (~ 10^-44 s)
>
> Practically we know that we all make simplifications. I can say I bought a
> Corolla from Luc at 2011-04-12 13:45 as an instant. I am not detailing how
> many seconds it took to sign the papers and hand over the key, but another
> asserter (using different activities) might do so, and would classify his
> Purchase activity to last from 13:44:42 to 13:45:36, ie. a
> time:ProperInterval.
>
> A third asserter might even disagree with the (instant) boundaries of that
> interval, because although I did finish pronouncing "OK" at 13:44:42.000,
> Luc could not hear so until 13:44:42.006 due to the speed of sound.
>
> So as different accounts (and this might be a big argument for doing
> subprocesses as nested accounts) can have different granularity, "instant
> events" grow to activities with duration. But we can't cater for the quantum
> physicist at all granularities.
>
> Note: The WG which proposed time.owl *draft* has long gone, I don't know if
> the document can be adopted by another WG for such clarifications.
>
> On Nov 29, 2011 6:22 PM, "Simon Miles" <simon.miles@kcl.ac.uk> wrote:
>>
>> Hi Stian,
>>
>> Thanks for the clarification. That's fine, but the text does still say
>> that instants can have "no interior points". From what you say, that
>> is simply something that can't be enforced, and we distinctly would
>> not want it to be, so it seems misleading to explain it that way.
>> Maybe what is meant is "no interior points for this asserter at this
>> moment", but that doesn't seem too helpful.
>>
>> Thanks,
>> Simon
>>
>> On 29 November 2011 16:39, Stian Soiland-Reyes
>> <soiland-reyes@cs.manchester.ac.uk> wrote:
>> >
>> > On Nov 24, 2011 3:13 PM, "Simon Miles" <simon.miles@kcl.ac.uk> wrote:
>> >
>> >> Maybe a philosophical point, but is an Instant, as referred to in Sec
>> >> 3.1.4 and subclassed to Time in 3.1.10.1, really helpful in provenance
>> >> data? It is defined as having "no interior points", but can one
>> >> asserter ever know that what they refer to as an instant will not need
>> >> to be decomposed by a future asserter?
>> >
>> > As we are using time.owl here you are free to assert times using what
>> > granularity and time.owl properties fit, for instance only minutes.
>> >
>> > time.owl does not provide any precision notion (you could add a cusbtom
>> > one), but neither does it say much about time equivalense for instances.
>> > (interval have lots of time relations)
>> >
>> > So I believe there would be no conflicts, although in general it would
>> > make
>> > more sense to talk about intervals between different activities, etc.
>>
>>
>>
>> --
>> Dr Simon Miles
>> Lecturer, Department of Informatics
>> Kings College London, WC2R 2LS, UK
>> +44 (0)20 7848 1166
>>
>> Provenance in Agent-mediated Healthcare Systems:
>> http://eprints.dcs.kcl.ac.uk/1273/
>>
>



-- 
Dr Simon Miles
Lecturer, Department of Informatics
Kings College London, WC2R 2LS, UK
+44 (0)20 7848 1166

Modelling the Provenance of Data in Autonomous Systems:
http://eprints.dcs.kcl.ac.uk/1264/
Received on Friday, 2 December 2011 17:32:12 UTC

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