Re: To spec editors - regarding possibly redundant rdf:type definition of containers in examples

Nice bait Arnaud ... pass.

I don't personally care if LDP chooses to relax the inferencing rule to 
the degree that (strawman!) non-leaf types defined by LDP or LDP's 
normative references can be inferred.

We just need to all be clear on and comfortable with the effects:

1:   The harder the rules are to describe, the more arbitrary it looks and 
the more complex/error-prone/comment-generating it is.  As in an outsider 
lamenting: what makes LDP's types so special?  Aside from "it's our spec, 
nyeh" what's our answer?  I had to insert 'non-leaf' to keep ldp:Basic... 
et al. in the representations, so it's already "not what my mother would 
expect" as one of our execs likes to say.

2:   "A hairbreadth's difference...."    We'd need consensus on the exact 
scope.  We know how much fun getting there is.  I'm not sure that "types" 
is the right scope; if you squint, you'll see that we have either skirted 
the edge of the non-inferencing rule in other places or outright violated 
it.  I'm thinking of cases where the goal and proposal was explicit ("keep 
the simple case simple"), so you get BC's that Must behave like they have 
insertedContentRelation=MemberSubject, but LDP imposes no requirement to 
materialize that triple.

3:  It's not about what is stored (which technically we should not care 
about, BTW), it's about what is transferred (representations) and how 
"classes"/"sets" of those are described (query).  E.g., if we allow 
ldp:Container (et al.) to not be materialized in representations, can 
clients usefully query for it?  If the query results are going to vary by 
implementation, because some impls "logically add" ldp:Container to graphs 
"but don't store it" and others do neither, debatable if that's "useful"; 
it's certainly not widely interoperable.


One way to think about removing the "extras" is as an optimization.  The 
first version need not be fully optimized ... if that's not true, let's go 
back to the I-word.


Best Regards, John

Voice US 845-435-9470  BluePages
Tivoli OSLC Lead - Show me the Scenario




From:   Arnaud Le Hors/Cupertino/IBM@IBMUS
To:     Cody Burleson <cody.burleson@base22.com>, 
Cc:     Linked Data Platform WG <public-ldp-wg@w3.org>
Date:   02/10/2014 05:07 PM
Subject:        Re: To spec editors - regarding possibly redundant 
rdf:type  definition of  containers in examples



Well, I'm glad you brought this up. I said the same thing to Steve and 
John. I think they are reading too much into the clause about not 
requiring inferencing. 

I don't read this clause as meaning we can't expect a client to know that 
an ldp:BasicContainer is an ldp:Container and that therefore the latter 
doesn't need to be materialized.
--
Arnaud  Le Hors - Software Standards Architect - IBM Software Group