Re: A proposal for clarifying the definitions of datatype maps, take II

On Wed, Jul 29, 2009 at 6:30 PM, Ian
Horrocks<ian.horrocks@comlab.ox.ac.uk> wrote:
> Alan,
>
> Getting rid of this MUST in Conformance is one of the main purposes of the
> cleanup -- the MUST was an ugly sticking plaster that was needed only in
> order to cover up the fact that Syntax didn't sufficiently constrain the OWL
> 2 Datatype map. Syntax now does this. As a result, we no longer need the
> sticking plaster.

There were three aspects of how the constraints needed to be made
clear. First, it needed to be tightened so that it was clear that
additional datatypes beyond a certain set pushed it out pushed it out
of DL. Second it needed to be clarified that a minimum set of
datatypes needed to be supported by OWL DL tools, and third it needed
to be made clear that we base our definitions on the established
external specifications such as XML Schema.

I believe the changes best served the first and third, but make it
hard to determine the second, which is why the removal of the MUST,
which directly spoke to the second, caught my attention. I see no harm
whatsoever in making the statement in conformance, so I don't
understand what the objection is. People will look to conformance to
understand issues like this.

>
> Conformance now simply says that conformant systems have to support the
> language as defined in the spec -- which is clearly what it always should
> have said. Of course some analysis is required in order to understand the
> spec, but this is true for all parts of it and not just datatypes.

Yes, but we don't have to make it harder for them. I agree that we
should not make a habit of having verbatim repeat elements in one
document and another. However, this is not the case on this issue - we
are simply clarifying something that might otherwise need a rather
careful read to get otherwise, in the place where one expects to find
such clarifications. (Actually we would be not removing a
clarification).

The proposal Boris made was to clarify/move things around. Your
summary: "What I understand is that you are just suggesting moving
things around to make docs more consistent". The MUST was not moved
from conformance to Syntax. Rather it was removed and a need to make
an inference was left in its place (i.e. it is less clear). Still, on
reading, I find it difficult to construct an argument that tools MUST
support all the datatypes, given what is in syntax now. So I don't see
the changes as completely meeting the proposal, despite me agreeing
that otherwise they improve things.

-Alan

p-Alan
Best,
Alan

>
> On 29 Jul 2009, at 17:22, Alan Ruttenberg wrote:
>
>> On Tue, Jul 28, 2009 at 6:39 PM, Ian
>> Horrocks<ian.horrocks@comlab.ox.ac.uk> wrote:
>>>
>>> The WG (including Ivan and Sandro) discussed and resolved on this already
>>> --
>>> see [1]. What I was reporting was simply the completion of the relevant
>>> action [2]. The rationale is that the change is only editorial, because
>>> we
>>> didn't change anything w.r.t. the overall spec -- the MUST simply moved
>>> from
>>> Conformance to Syntax.
>>
>> Hi Ian,
>>
>> But my point is that the MUST was lost in the process, as is confirmed
>> by Boris' email. While it may be the case that the effect is the same,
>> it now requires some inference to arrive at the conclusion.
>>
>> To be honest I would prefer if we had a MUST somewhere, even if it
>> seems redundant on careful analysis.
>>
>> -Alan
>>
>>
>>>
>>> My understanding is that we will report all such changes in the PR
>>> versions
>>> of the documents.
>>>
>>> Let me know if you think we need to do something more.
>>>
>>> Regards,
>>> Ian
>>>
>>> [1] http://www.w3.org/2007/OWL/meeting/2009-07-01#resolution_2
>>> [2] http://www.w3.org/2007/OWL/tracker/actions/344
>>>
>>>
>>>
>>> On 28 Jul 2009, at 22:20, Jim Hendler wrote:
>>>
>>>> I hate to be a stickler for process, but this seems like it is slightly
>>>> more than just a bug fix (esp. as it, appropriately, removes a MUST
>>>> clause)
>>>> - to be clear, I think this is a good change and I think it does not
>>>> invalidate a move to CR, but we should ask our team reps (Ivan and
>>>> Sandro)
>>>> to make sure we are compliant with process and do whatever notification
>>>> we
>>>> need to do (if any) so that we can move ahead -- I think it may just
>>>> need to
>>>> be added as a note in the CR documentation (that we made this clarifying
>>>> change), but at this late date let's be sure to dot our i's and cross
>>>> our
>>>> t's (as the expression goes)
>>>>  -JH
>>>>
>>>>
>>>> On Jul 28, 2009, at 6:09 AM, Ian Horrocks wrote:
>>>>
>>>>> It might be worth adding that this was exactly the motivation for the
>>>>> clarification/cleanup. As things stood before, the concept language was
>>>>> precisely defined in Syntax/Profiles, and Conformance simply said that
>>>>> conformant systems had to support the concept language as defined in
>>>>> those
>>>>> documents; in contrast the datatype language was relatively loosely
>>>>> defined
>>>>> (or at least allowed for some variability), and Conformance "fixed"
>>>>> this by
>>>>> stating that conformant systems must support all OWL 2 datatypes. This
>>>>> was
>>>>> clearly undesirable -- important parts of the language specification
>>>>> should
>>>>> not be "hidden" in Conformance.
>>>>>
>>>>> The result of the clarification is that Syntax/Profiles now precisely
>>>>> define the datatype part of the language just as for the concept part.
>>>>> Conformance can thus simply say that conformant systems must support
>>>>> the
>>>>> language as defined in Syntax/Profiles.
>>>>>
>>>>> Ian
>>>>>
>>>>>
>>>>> On 28 Jul 2009, at 07:43, Boris Motik wrote:
>>>>>
>>>>>> Hello,
>>>>>>
>>>>>> While it is true that this sentence was removed, I don't think that
>>>>>> anything has
>>>>>> been lost from the normative point of view.
>>>>>>
>>>>>> The Syntax document now defines in Section 4 the OWL 2 datatype map as
>>>>>> a
>>>>>> fixed
>>>>>> set of datatypes; then, in Section 5 it says that people can use these
>>>>>> datatypes
>>>>>> in OWL 2 ontologies. Datatypes are now just like any other construct:
>>>>>> they are a
>>>>>> fixed part of the language. Saying something like "an OWL 2 tool must
>>>>>> support
>>>>>> all OWL 2 datatypes" is thus tantamount to saying "an OWL 2 tool must
>>>>>> support
>>>>>> all OWL 2 class constructors".
>>>>>>
>>>>>> The sentence you refer to has been introduced because things have not
>>>>>> been like
>>>>>> this earlier: the set of datatypes was not fixed and we initially
>>>>>> allowed for a
>>>>>> pick-and-mix approach. Since this is now completely gone from all
>>>>>> parts
>>>>>> of the
>>>>>> Syntax document (as well as the other documents), I really don't think
>>>>>> anything
>>>>>> special needs to be said about the support for datatypes: they need to
>>>>>> be
>>>>>> supported in their entirety just like any other part of the language.
>>>>>>
>>>>>> Regards,
>>>>>>
>>>>>>       Boris
>>>>>>
>>>>>>> -----Original Message-----
>>>>>>> From: public-owl-wg-request@w3.org
>>>>>>> [mailto:public-owl-wg-request@w3.org] On
>>>>>>> Behalf Of Alan Ruttenberg
>>>>>>> Sent: 28 July 2009 04:30
>>>>>>> To: Ian Horrocks
>>>>>>> Cc: OWL 1.1; Boris Motik
>>>>>>> Subject: Re: A proposal for clarifying the definitions of datatype
>>>>>>> maps, take
>>>>>>> II
>>>>>>>
>>>>>>> I may have missed something, however it appears that these changes,
>>>>>>> while clarifying the meaning of the datatypes in the OWL 2 Datatype
>>>>>>> map, also remove a strong constraint - namely that OWL 2 DL tools
>>>>>>> MUST
>>>>>>> support all the types in that datatype map.
>>>>>>>
>>>>>>> In particular:
>>>>>>>
>>>>>>> "OWL 2 tools <em title="MUST in RFC 2119 context"
>>>>>>> class="RFC2119">MUST</em> support the OWL 2 datatype map described in
>>>>>>> the rest of this section. "
>>>>>>>
>>>>>>> has been removed.
>>>>>>>
>>>>>>> I don't believe that Boris' original note suggested this would be the
>>>>>>> case.
>>>>>>>
>>>>>>> I'd appreciate some clarification on this matter.
>>>>>>>
>>>>>>> Thanks,
>>>>>>> Alan
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> On Sun, Jul 26, 2009 at 4:52 PM, Ian
>>>>>>> Horrocks<ian.horrocks@comlab.ox.ac.uk> wrote:
>>>>>>>>
>>>>>>>> As you will recall, the WG approved Boris's proposal during the 1st
>>>>>>>> July
>>>>>>>> teleconf [1]. Completing the necessary work has taken a while --
>>>>>>>> entirely my
>>>>>>>> fault for being slow to do the necessary work on Conformance.
>>>>>>>>
>>>>>>>> To summarise, Boris has clarified the definition of datatypes and
>>>>>>>> the
>>>>>>>> OWL
>>>>>>>> datatype map in Syntax. As a result, Conformance no longer needs to
>>>>>>>> specify
>>>>>>>> constraints on datatypes and the datatype map (e.g., that conformant
>>>>>>>> tools
>>>>>>>> must use the OWL 2 datatype map) -- the datatypes that can occur in
>>>>>>>> (profile) documents and that must be supported by (profile) tools
>>>>>>>> are
>>>>>>>> now
>>>>>>>> explicitly defined in Syntax and Profiles. The relevant diffs are:
>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> http://www.w3.org/2007/OWL/wiki/index.php?title=Syntax&diff=24783&oldid=24704
>>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> http://www.w3.org/2007/OWL/wiki/index.php?title=Syntax&diff=24850&oldid=24798
>>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> http://www.w3.org/2007/OWL/wiki/index.php?title=Conformance&diff=24942&oldid=2
>>>>>>> 4877
>>>>>>>>
>>>>>>>> Please let us know ASAP if you have any comments w.r.t. these
>>>>>>>> changes.
>>>>>>>>
>>>>>>>> Regards,
>>>>>>>> Ian
>>>>>>>>
>>>>>>>> [1] http://www.w3.org/2007/OWL/meeting/2009-07-01#resolution_2
>>>>>>>>
>>>>>>>>
>>>>>>>> On 29 Jun 2009, at 14:33, Boris Motik wrote:
>>>>>>>>
>>>>>>>>> Hello,
>>>>>>>>>
>>>>>>>>> In April I've sent around the following e-mail, in which I've
>>>>>>>>> proposed to
>>>>>>>>> clarify certain definitions surrounding datatype maps:
>>>>>>>>>
>>>>>>>>> http://lists.w3.org/Archives/Public/public-owl-wg/2009Apr/0454.html
>>>>>>>>>
>>>>>>>>> Please refer to my original e-mail for the details; in short, the
>>>>>>>>> idea is
>>>>>>>>> to
>>>>>>>>> remove certain discrepancies between Conformance and the rest of
>>>>>>>>> the
>>>>>>>>> documents,
>>>>>>>>> with Conformance being taken as a guideline.
>>>>>>>>>
>>>>>>>>> I haven't pushed this forward earlier because we were getting ready
>>>>>>>>> to go
>>>>>>>>> into
>>>>>>>>> CR. Since we've successfully reached that milestone, now seems like
>>>>>>>>> a
>>>>>>>>> perfect
>>>>>>>>> time for improving the spec. Therefore, unless someone objects, I
>>>>>>>>> would
>>>>>>>>> make a
>>>>>>>>> few editorial changes to the spec and inform the WG of the outcome.
>>>>>>>>>
>>>>>>>>> Regards,
>>>>>>>>>
>>>>>>>>>     Boris
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>
>>>> We choose to go to the moon in this decade and do the other things, not
>>>> because they are easy, but because they are hard - John F. Kennedy, Sept
>>>> 12,
>>>> 1962
>>>>
>>>> Prof James Hendler
>>>>  http://www.cs.rpi.edu/~hendler, @jahendler, twitter
>>>> Tetherless World Constellation Chair
>>>> Computer Science Dept
>>>> Rensselaer Polytechnic Institute, Troy NY 12180
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>
>>>
>
>

Received on Thursday, 30 July 2009 01:46:02 UTC