W3C home > Mailing lists > Public > www-qa-wg@w3.org > August 2002

Re: Spec GL: "explictly state" CP

From: Lofton Henderson <lofton@rockynet.com>
Date: Fri, 02 Aug 2002 15:16:03 -0600
Message-Id: <>
To: David Marston/Cambridge/IBM <david_marston@us.ibm.com>, dominique HazaŽl-Massieux <dom@w3.org>
Cc: www-qa-wg@w3.org

Like Dom, I have always been a little uncomfortable with encumbering all 
specifications with having to make the negative disclaimer about each 
unused DoV.  At the same time, believing that there is value that each DoV 
be explicitly considered, and explicitly adopted or rejected.

David's idea is interesting.  In fact, our "Spec ExTech" could provide a 
handy 8-row table (or 9-row, for "other") -- "yes", "no", "n/a" (whatever 
that means) -- for cut-paste into your specification's conformance clause.

For further discussion, is anyone interested to draft prototype text to 
show what the first 1-2 ckpts of GL.3 (for example) would look like?  And 
what new checkpoint 10.x would look like?


At 12:45 PM 8/2/02 -0400, David Marston/Cambridge/IBM wrote:
>Dominique writes about:
>"Otherwise, explicitly state that [this dimensions] is not supported."
> >I'm not sure we want to encumber all of the W3C Spec with a mention of
> >no support for all these dimensions. It would probably be better to
> >propose a generic way of stating that any other dimension of
> >variability not explicitly allowed is not supported [but without using
> >the very QA-only expression of "dimension of variability".]
>One possible approach is to remove that negative checkpoint from each
>dimension, but have a pointer in each across to something in GL 10:
>"If [this dimension] is not used, make an explicit statement of non-use
>per the provisions of Checkpoint 10.x." Then, part of Ck 10.x says:
>"For each of the following specification techniques that is not used,
>state that fact explicitly: multiple classes of product, profiles,
>modules, degrees of conformance, deprecation, levels, discretionary
>choices, extensibility, and open-ended discretion to support languages,
>locales, etc." (rough wording to convey the general idea)
>This way, a person bent on implementing or testing the spec can start
>from the table of contents, navigate to conformance verbiage, and find
>out what variability exists, as well as those dimensions that are
>designated as invariant.
>.................David Marston
Received on Friday, 2 August 2002 17:12:59 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:14:28 UTC