Re: review of rdf:text, dated 2008-11-04

>> I reviewed the current draft of the rdf:text specification [1].
>> I subdivided my comments into criticism on the content, criticism on the
>> structure, errors in the document, and editorial issues.
>>
>> Criticism on the content
>> ====
>> - to assure maximum compatibility with current and future versions of
>> XML schema datatypes, the string parts of both the lexical and value
>> space should be based on the respective spaces of the XML schema
>> datatype string.
>> - the set of characters is finite, and thus it cannot be assumed that it
>> is infinite. The problem that some OWL 2 implementations might have some
>> issue with the finiteness of this set is of no concern to this datatype
>> per se. In fact, the XML schema string datatype is based on a finite set
>> of characters, and so OWL 2 implementations will run into problems with
>> this datatype.
>> If there is really a problem to be expected with implementations of OWL
>> 2, it should be dealt with in the OWL 2 specification, and not the
>> specification of this datatype.
> 
> I think you are not quite grasping the issue here. (I do prefer finite
> alphabets myself, fwiw.) The point is how to design the type so that it
> is extensible to additional characters that will definitely be added (by
> unicode). Note that problems along these lines have already occurred in
> XML land. I don't think we can *merely* punt on this.

What problems are there in XML land?
In any case, like XML, this definition relies on ISO/IEC 10646, which
has provisions for extensions. It seems to me that this is the
appropriate way to deal with extensibility; it's not necessary to define
our own mechanism.
In general, I think that this datatype should be based on the XML schema
string datatype, and if there are problems with extensibility, they
should be solved in XML schema.

> (And the problem is that future changes will change the meaning of some
> ontologies. I presume that this will be true for some RIF rulesets if
> you have the appropriate facets and builtins.)

We don't have such problems in RIF, because we don't allow built-ins in
rule head.
Further, if future changes in data types potentially pose a problem to a
particular language, the specification of that language should deal with
this problem. I suspect that OWL 2 does something like that for the
string data type.


best, Jos

> 
> Cheers,
> Bijan.
> 

-- 
Jos de Bruijn            debruijn@inf.unibz.it
+390471016224         http://www.debruijn.net/
----------------------------------------------
No one who cannot rejoice in the discovery of
his own mistakes deserves to be called a
scholar.
  - Donald Foster