Re: RDF/XML shorthand for RDF reification

On Jun 26, 2008, at 5:53 AM, Bijan Parsia wrote:

> On 25 Jun 2008, at 22:48, Alan Ruttenberg wrote:
>
>> On Jun 25, 2008, at 3:50 PM, Peter F. Patel-Schneider wrote:
>>
>>> The point I was making was that using the shorthand results in the
>>> reification node having a real name, i.e., not being a blank node,  
>>> which
>>> messes up lots of things, including parsing and semantics.
>>
>> Just for the record, how will it mess up parsing and semantics. All  
>> of our use of reification is for axioms. As I see it, the  
>> difference would be whether axioms had names in OWL Full (on the  
>> semantics side) and on the parsing described in table 6 and 17,  
>> which seem like they could be adjusted to used named instead of  
>> blank nodes.
>
> Here's a possibility:
>
> Suppose I serialize the same ontology with annotations on axioms  
> twice using two different serializers, S1 and S2. S1 sorts the  
> axioms lexicographically, then generates names for the axioms  
> starting from a seed and prefixing it with the xml:base, plus some  
> urn prfix. S2 also sorts lexicographically, but ascending (whereas  
> S1 is descending) and uses the same genname function.
>
> Now suppose I merge (or import) these two ontologies. It seems that  
> I would get some potentially strange results.
>
> Naming axioms should done with *extreme* care.

Thanks for the example.
I agree that care would need to be taken. It seems that what needs to  
be avoided are collisions. Allocating unique names is a fairly common  
practice these days - a combination of mac address, system time in  
milliseconds , and a reasonably sized random salt should be sufficient  
to ensure uniqueness.

What about the other way around - the "same" axiom having multiple  
differently named reifications - this would seem to be analogous to  
the current bnode case?

>
>
> Cheers,
> Bijan.
>