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

Re: rewrite of TOC checkpoints.

From: Lofton Henderson <lofton@rockynet.com>
Date: Mon, 04 Nov 2002 08:39:05 -0700
Message-Id: <>
To: Alex Rousskov <rousskov@measurement-factory.com>
Cc: www-qa@w3.org

At 06:32 PM 11/3/02 -0700, Alex Rousskov wrote:

>On Sun, 3 Nov 2002, Lynne Rosenthal wrote:
> > In Guideline 13, add:
> >
> > Checkpoint:  Provide a way to find conformance information
>If checkpoint "title" is not normative/testable text, I would suggest
>saying some thing like "easy way" or "fast way" to make the intention

I agree with this.  Although we had decided to remove untestable 
(subjective) words from checkpoint titles, that is not strictly necessary 
now that we are making a clear separation of "fulfillment criteria" (test 
assertions) from the title.

> > To fulfill this checkpoint, a specification MUST provide at least one
> > navigation mechanism that allows the reader to locate all
> > conformance-related information in the specification.
>Two problems/concerns:
>         - scanning a spec word-by-word can be considered
>           "navigation"; searching the document for word
>           "conformance" is a navigation mechanism; thus,
>           every text-based spec will meet your requirement

Perhaps a definition of navigation mechanism would help, or a limitation on 
what is considered valid.  Something with some sort of directness and 
linkage, as opposed to a instruction to the reader to "grep on all MUST 

(Hmm... having said that, a smart "complete-sentence" grep with line 
numbers and maybe even links -- as opposed to the standard unix single-line 
dumb 'grep' -- doesn't sound bad.  I.e., the navigation mechanism could be 
some extra-document process that extracts and presents one or more 
information-specific tabulations or TOCs.)

>         - if conformance clause requires that all MUSTs are
>           supported (for example), does "all conformance-related
>           information" include location of every MUST? If yes,
>           that may be impossible to index in a usable way. If no,
>           "all related" should be defined to exclude individual
>           checkpoints (like "all MUSTs").

The latter interacts with the test assertion checkpoints, which are still 
being revised.

Received on Monday, 4 November 2002 10:38:43 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:40:30 UTC