Re: PROV-ISSUE-496: ivan's feedback on prov-n LC [prov-n]

On 16/10/2012 23:37, Luc Moreau wrote:
>> [[[
>> The scope of a namespace declaration directly occurring in a document is the
>> document itself, exclunding the bundles it may contain
>> ]]]
>>
>> is that, in case of
>>
>> document
>>    prefix a <http:...>
>>    ...
>>    bundle abcde
>>      ... a:something ...
>>    endBundle
>> endDocument
>>
>> the usage of 'a:something' is illegal. I am a bit surprised by this
>> definition; usually (eg, in the trig-in-preparation) the prefix declarations
>> given on the top are valid for the whole document, but you may have had good
>> reasons for this restrictions that was discussed before I joined the group.
>> Ie, I will not object to that, but just wanted to be sure that we do have the
>> same understanding.
> Your interpretation is correct.
>
> There was a very *strong* push back by several WG members against any notion of
> scope related to bundles.  So, we adopted
> this solution that makes bundles in prov-n totally self-contained.
> The downside is that we can't inherit a prefix from a document, instead we have
> to replicate the prefix declaration.
> The good thing is that bundles are totally independent, and this fact is
> exploited in prov-constraints.

As one who was very concerned about scope in bundles, I think there may be a 
confusion here.

My concern was with scope of *identifiers*, specifically URIs, that would lead 
to the possibility of the same URI denoting different things in different 
bundles.  This would make it problematic to merge RDF representations of 
information from different bundles (under the reasonable expectation that URIs 
in PROV-DM/PROV-N are preserved in the RDF representations).

The scoping of prefix declarations is a different matter.  Prefixes are simply a 
syntactic mechanism for constructing identifiers.  As such, I see no problem 
with scoping of prefix declarations as long as the resulting identifiers are 
understood to be not scoped to bundles.  This would be very similar to the use 
of XML namespaces to scope namespaces used for properties in RDF/XML.

That said, I have no overwhelming problem with the resolution proposed, though I 
can see that it could lead to what some might regard as proliferation of prefix 
declarations.  If PROV-N is expected to be predominantly hand-crafted, then I 
would be inclined to allow visibility of outer prefix declarations in nested 
bundles as a matter of convenience for writers.

#g

Received on Sunday, 21 October 2012 12:50:45 UTC