Re: grddlonrdf tests

@@phpht. I dunno if this message makes any sense; I haven't really 
finished it; I'm sending it anyway....

Chimezie Ogbuji wrote:
> On Wed, 7 Mar 2007, Dan Connolly wrote:
>> The short description of this tests says "/An implementation only has 
>> to produce one of these three/."
>> (1) that's not true, i.e. can't be justified from the spec. producing 
>> none is consistent with the spec.
>
> Hmm.. that's not my understanding and is the very reason why I was 
> concerned about this particular scenario in our last telecon: i.e., 
> how can a piece of software which doesn't produce any GRDDL results ( 
> when there should be at least one ) be considered a GRDDL-aware-agent 
> by the current definition?
A piece of software that handles the glean-hcal an *no other 
transformations*, by policy, is
a GRDDL-aware agent, but it won't compute the GRDDL results in the case 
of this test.

(by the way: it's not that "there should be a result"; there _is_ a 
result. The only question
is whether various bits of software find/compute them or not.)

>
> Barring an explicit choice to ignore a nominated transform due to " 
> the agent's capabilities, local security policies
It's exactly that choice that makes computing zero results consistent 
with the spec.

> and possibly user/client intervention." it would *not* be a 
> GRDDL-aware agent.  That's my interpretation of section 7.
>
>> (2) "an implementation" isn't a term we've defined; "GRDDL-aware 
>> agent" is the conformance label
>> (3) let's please not refer to implementations from the test 
>> descriptions at all. The tests are there
>> primarily to clarify the language design. They are also intended to 
>> help developers develop interoperable
>> code, but I'd rather leave that out of the test descriptions altogether.
>>
>> Also, re "there are /three distinct and equally valid output graphs 
>> for this test/ for this document"
>> (1) "equally valid" is a little tricky to justify from the spec, 
>> given the SHOULD stuff we recently
>> put in there re GRDDL-aware agents.
>
> Well if equally valid means they are 'GRDDL results', it is pretty 
> clear since we describe the only means by which GRDDL results can be 
> computed. The SHOULD stuff in section 7 allows the agent to make 
> determinations as to which transform to apply, with the possible 
> consequence (evil, untested hook?) of causing it to no longer be a 
> compliant GRDDL-aware agent if such determinations result in not 
> applying *any* transformations.


-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/

Received on Wednesday, 7 March 2007 16:52:50 UTC