Re: PROPOSAL to close ISSUE-16: RDF Collections


Hearing your clarifications re the below leads me to believe that:

  - Lists/Collections are often needed

  - The variation in usage of Collection vs List is more determined by 
the mediatype, for instance List is more popular with N3/Turtle (easier 
to write), Collection with RDF/XML (easier to write), Neither with RDFa 
(complex to write).

Personally I have many uses for Lists specifically, but that's I feel 
that whether you consider a list as ordered or not should be context 
specific - much like a common array in programming which serves as a 
single structure for both indexed lists and unordered collections - thus 
I find myself with little use for Collections.

Also, I did not find your proposal too complex, rather it appeared to 
provide a single easy-to-use syntax (::member) for both Lists and 
Collections, and this really appealed to me.

The only issue I had was down to time, if there was a way to get your 
proposal to a level where consensus could be achieved in time then I'd 
be fully behind it and help where I can. Perhaps if we fail to reach LC 
on Thursday then we could try and get this done?



Ivan Herman wrote:
> Again for the records: I have to point out some mistake in the argumentation. 
> On Oct 18, 2010, at 23:23 , Manu Sporny wrote:
> [snip]
>> 1. Hardly anybody uses them on the OpenWeb - hit the billion triples
>>   challenge data and you will only find a handful of rdf:Seq, rdf:Bag,
>>   and rdf:List triples.
>> 2. RDF Collections are used heavily in OWL and XMP.
> These two points contradict to one another:-(
>> 3. People tend to fall back to using things like ex:position to provide
>>   lists. Unordered lists are usually expressed as a collection of
>>   properties with different values. People are modeling collections
>>   in different ways - there is no clear pattern.
> There is, the issue is with SPARQL 1.0 rather than with anything else. Collections in RDF are fine semantically (after all, they are Lisp lists in RDF, nothing else), but SPARQL has difficulties querying them (although SPARQL 1.1 will make the situation a bit better).
>> 4. Expressing linked lists is something that many people don't need to
>>   do, no need for the added complexity in RDFa. Why do we need to
>>   add a feature that hardly anybody is going to use?
> This is, unfortunately, a very circular argument. Adding collections (lists) to an RDFa file is, at the moment, extremely complicated and ugly. So nobody uses them...
>> 5. It's not the RDFa WG's job to figure out how to best express
>>   collections in RDF - it should be part of the RDF Next WG's work.
> I do not agree with the argument that this is RDF Next WG's work (and I do not think it will be part of the charter). Collections is well specified in RDF and there is no reason to come back to them right now. 
> I accept the argument that the design I have put forward is complex, and that there is no consensus in the group to move forward with it at this point. Ie, I do not object to the conclusion of the issue, but I do not really think the arguments above are the right one.
> Ivan
>> Opinions on each item above varied greatly, but the thoughts at the end
>> of the discussion were that we were not going to be able to reach
>> consensus on how to express lists in RDFa much less how we would do it.
>> This proposal asserts that the complexity of designing a mechanism for
>> expressing collections in RDF is very large and that there is no clear
>> path forward that would gain the consensus of the RDFa WG, therefore the
>> issue should be closed and referred to the RDF Next WG for further
>> consideration when it starts up.
>> Please comment before Thursday, October 21st at 13:00 UTC if you object
>> to this proposal. If there are no objections by that time, this issue
>> will be closed. If there are objections, the RDFa Working Group will
>> perform a straw-poll and decide whether or not to close the issue before
>> entering Last Call.
>> -- manu
>> -- 
>> Manu Sporny (skype: msporny, twitter: manusporny)
>> President/CEO - Digital Bazaar, Inc.
>> blog: Saving Journalism - The PaySwarm Developer API
> ----
> Ivan Herman, W3C Semantic Web Activity Lead
> Home:
> mobile: +31-641044153
> PGP Key:

Received on Tuesday, 19 October 2010 10:05:55 UTC