Re: Input sought on datatyping tradeoff

>From: pat hayes <phayes@ai.uwf.edu>
>Subject: Re: Input sought on datatyping tradeoff
>Date: Wed, 17 Jul 2002 12:57:04 -0500
>
>>  >I feel somewhat like the man who was asked if he was to be condemned
>>  >to be executed whether he would like scrambled eggs or pancakes for
>>  >breakfast before the execution.  After he replied ``Pancakes, of
>>  >course'' he was told that then it would certainly be all right to
>>  >execute him.
>>  >
>>  >We are being asked to pick between two choices that are only on the
>>  >table because of various constraints imposed by not only XML and RDF
>>  >but by the members of the RDF Core Working Group.  I have this
>>  >uneasy feeling that the results of the poll will then be used to
>>  >justify the constraints that are being discussed in the working
>>  >group.  In this sort of situation it is only right to reject the
>>  >poll.
>>
>>  Look, the WG is only trying to get feedback from as wide as possible
>>  a section of the user community in order to help it make a decision.
>>  It has tried to make the decision itself and has found itself
>>  deadlocked on which way to go. Under these circumstances, asking for
>>  wider feedback seems a reasonable way to proceed. I see no reason why
>>  you or anyone should feel uneasy at this decision being made in this
>>  way. A decision has to be made, one way or the other.
>
>Well the message sent out was not
>
>      We have this way writing literals that is causing us problems because
>      we cannot determine what it should mean.  We have two different
>      options, one of which [describe how the tidy option works and what
>      implications it has] and the other of which [describe how the untidy
>      option works and what implications it has].  Which of these two ways
>      do you think is better, or do you have another way of looking at
>      literals that gets around this problem?
>
>It was
>
>      We have this way of writing literals that is causing us problems.  We
>      have to make a choice between [test case A] and [test case D].  Pick A
>      or D, but not both, and tell us why.
>
>If the WG had wanted feedback there was much better ways of requesting it.
>

I sometimes think that if I were to lie on the ground and invite you 
to walk over me, you would object to the color of my shirt. Look, 
Brian was trying to focus the issue. As you know, these matters are 
long and involved and have many ramifications which are hard to 
explain. For example, there isn't a single 'untidy option' ; there 
are at least three distinct untidy options, at least two notions of 
'tidy' (syntactic and semantic) and two directly incompatible views 
on how best to think of literal-to-value maps (as a property or as 
part of the denotation mapping). There are differing views on the 
relative importance of monotonicity, global/local datatyping, 
uniqueness of reference, context-independence of reference, whether 
or not we can meaningfully interpret RDF datatypes as canonical, and 
so on and on. What we most definitely do not want to do is open up 
this entire area for a public debate, particularly when time is so 
short.

We have found in general (ie not just in regard to this particular 
issue) that making up test cases and asking people to decide which 
way they would like them to go is often a good way to focus 
discussion, and indeed these particular test cases have been the 
focus of many of our discussions, and have recurred again and again. 
If anyone is interested in reading the entire discussion in all its 
complexity they are free to read the email archive, if they feel they 
have the time and the inclination.

>  > Your response presumably means that you refuse to provide any input.
>>  I take it, therefore, that you will not complain when your views,
>>  whatever they happen to be, are not taken account of when the
>>  decision is made.
>>
>>  Pat
>
>I have provided lots of input, both previously and also in response to the
>request, part of which was to the effect that the request is ill-formed.

It is not ill-FORMED. It may indeed reflect some background 
discussion that you do not like, but then we are not putting the 
entire subject up for public debate, only seeking feedback on one key 
issue.

>I have, moreover, provided two different ways of looking at literals in
>response to the request, as well as several other proposals for datatyping.

I know you have, but (1) the request wasn't for new proposals for 
datatyping - we have too many already - and (2) none of these ways 
are acceptable to the WG, for various reasons. The proposal in your 
recent 'brief' summary, for example, was considered and rejected by 
the WG on other grounds. (My own objection to it is that it does not 
provide for the use of literals without datatypes; another objection 
is that it is can be easily mapped into a notational variant on one 
of the existing proposals.) It is not appropriate to re-open this 
entire area of discussion at this point in the decision process.

One option we still have is to simply punt on this issue, and I think 
that the only way to usefully take advantage of your input would be 
to take it as a recommendation to punt.

Pat

-- 
---------------------------------------------------------------------
IHMC					(850)434 8903   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola,  FL 32501			(850)202 4440   fax
phayes@ai.uwf.edu 
http://www.coginst.uwf.edu/~phayes

Received on Wednesday, 17 July 2002 17:00:37 UTC