W3C home > Mailing lists > Public > public-grddl-wg@w3.org > April 2007

Re: Updated Test Document to reflect actions / discussion from last teleconference

From: Jeremy Carroll <jjc@hpl.hp.com>
Date: Tue, 17 Apr 2007 15:36:29 +0100
Message-ID: <4624DB6D.6000806@hpl.hp.com>
To: ogbujic@ccf.org
CC: GRDDL Working Group <public-grddl-wg@w3.org>

Hi Chime

This is looking good.

I have a few minor points about specific tests and wordings


a) Namespace loop

You ask whether the namespace loop tests violate the general policy 
concerning maximal grddl results.

They do.

My understanding was that the WG was minded that the '"SHOULD" produce 
GRDDL results' can reasonably be understood as allowing implementers 
making a fair effort to implement the spec, to not compute the maximal 
result in a document with this sort of self-dependence.

Hence, I am imagining that these four tests are
a) passed by explicit vote of WG, aware that this violates our own 
policy, and hence gives the reader a view of our interpretation of the 
b) have a small explanatory text to the effect that the approval of 
these tests reflect the WG's interpretation of the 'SHOULD' in the 
specification with regards to this input.

The following four tests, show different GRDDL results possible
for a particular self-referential input document.
To be consistent with the other tests, only the last of these,
the maximal result, would apply.
The Working Group approved all four tests on @@@Not yet approved@@@
to reflect an interpretation of the RFC  2119 keyword 'SHOULD' in
GRDDL specification section 7.
The Working Group, aware of the implications
of not producing all the GRDDL results in this case,
and aware of the difficulty of so doing,
do not discourage implementers from being satisfied with passing
any one of these four tests.

That's probably too wordy, I was referring to RFC 2119.

b) content negotiation

I am unclear as to why the two content negotiation tests are in the 
informative section - they look pretty normative to me.
I should produce a maximal result for these.

c) multiple representations test seems to have got lost.
I will re-propose it in the pending list.


d) picky point about tense
    "will" occurs in
The test will have a symmetric property


The tests which require a network connection will use absolute URIs

I suggest delete 'will' in both cases

[We're doing well when review is reduced to such picky points - it means 
the substance is fine!]

e) slightly less picky.
I wasn't comfortable with the phrasing of the following sentence:

However, for the purpose of running these tests in order to determine 
compliance, a GRDDL-aware agent with a security policy which does not 
prevent it from applying transformations identified by each test will 
produce the GRDDL result associated with each normative test.

The word I don't like is 'determine'.
The tests can show that some software is not a GRDDL-aware agent, but 
cannot show that it is.

I would prefer a phrasing that did not have a reading that a piece of 
software that passes all the tests is a GRDDL-aware agent.

However, for the purpose of defining these tests it was assumed that
the GRDDL-aware agent being tested is using a security policy which does 
not prevent it from applying the transformations identified in each test.
Such an agent will produce the GRDDL result associated with each 
normative test, except as specified immediately below.

Withdrawl of Previous Comment

On #alternative and #subsumes.

I believe the document, and the manifests, are fine without these 
features, and would be happy to see the two relevant paragraphs deleted.

However, I will argue no longer for their removal.
They have merely editorial force, and you are the editor.
Since this is essentially a matter of taste, I will support your 
judgement (at least at the formal level!).

In other words, I think we have discussed this point enough, and we 
should agree to differ - and so you win, by virtue of holding the pen!


Chimezie Ogbuji wrote:
> Per my part of the action [1] from the last teleconference, I updated
> the test document [2].  In particular I updated the section on multiple
> GRDDL results to reflect our decision to collapse tests with a single
> infoset/representation (but multiple GRDDL results) into the maximal
> result.  The approved tests have been updated appropriately.
> RCS file: /w3ccvs/WWW/2001/sw/grddl-wg/td/grddl-tests.html,v
> ----------------------------
> revision 1.19
> date: 2007/04/16 20:03:31;  author: cogbuji;  state: Exp;  lines: +61
> -33
> - updated multiple output section to clarify the 3 kinds of multiple
> output scenarios
> - removed background color for approved test links
> - added approval links for tests approved during 4/11 teleconference
> - Collapsed single infoset / representation multiple output tests into
> maximal result
> ----------------------------
> [1] http://www.w3.org/2007/04/11-grddl-wg-minutes.html#action06
> [2] http://www.w3.org/2001/sw/grddl-wg/td/grddl-tests

Hewlett-Packard Limited
registered Office: Cain Road, Bracknell, Berks RG12 1HN
Registered No: 690597 England
Received on Tuesday, 17 April 2007 14:37:04 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:39:11 UTC