Re: Answer to Ian Hickson: Formal vs prose language normativity

At 4:34 PM +0000 5/3/05, Ian Hickson wrote:
>On Tue, 3 May 2005, Karl Dubost wrote:
>>
>>  Original comment (issue 1049 [1])
>>  http://lists.w3.org/Archives/Public/www-qa/2005Jan/0014.html
>>
>>  Thank you for your comment, which the QA Working Group has accepted. 
>>  We have reworded the affected section as you recommended and it now
>>  reads
>>
>  > [...] there are possible overlaps between the prose and the formal
>>  language, in which case, it is important to define which one is the main
>>  point of reference in case of disjunction.
>
>This is actually exactly the opposite of what my comment said. In my
>opinion, when there is a conflict it means that one or the other of the
>formal language and the prose is incorrect. There is no guarentee about
>which one is the correct one.
>
>Whenever there is a conflict between prose and formal language, the
>working group must, IMHO, release errata fixing the problem. Saying that
>one overrides the other implies some sort of belief that errors will only
>creep into one and not the other, which is clearly not going to be the
>case. It also means that people will be discourages from reporting errors
>to the editors since errors would always be "resolved" (albeit effectively
>without working group supervision).
>
>I do not accept this resolution.

I wonder if we can converge this issue, yet.

<quote class= "changeFrom" cite=
"http://www.w3.org/TR/2005/WD-qaframe-spec-20050428/#formal-language-gp">

However, prose remains necessary to allow implementers to understand
the specification, as well as to express additional requirements the
formal language cannot express; this means that there are possible
overlaps between the prose and the formal language, in which case, it
is important to define which one is the main point of reference in
case of disjunction.

</quote>

<draft class="changeTo">

However, prose remains necessary to allow implementers to understand
the specification, as well as to express additional requirements the
formal language cannot express; this means that there are possible
overlaps between the prose and the formal language. In this case, if
the developers of the specification have a clear position on which
one is the main point of reference in case of conflict, this
precedence should be clearly stated in the document.

</draft>

<discussion>

In general, I would support Hixie's "tiebreaker rules considered harmful"
position.  On the other hand, I recently posted the following to explain
the tiebreaker position taken on another document in another group.

<blockquote>

[...]

This is actually a matter of art, not science.  Once over quickly:

None of the available schema notations, XSD, RELAX-NG, etc.
actually says everything that you want to specify.

Search for "Masayasu Ishikawa" at
http://www.w3.org/2004/03/plenary-minutes#Session4

Different groups put more attention on the schemas and decide
to make them normative or informative depending on how central
they were in the development process.  Since in our case we
consensed on the prose and derived the schemas after the fact,
we have told the reader that they are informative.  That doesn't
keep you from validating against the schemas as a check on both
the schemas and the examples.

But if your schema validation fails, you need to compare the
error message you get with the main body prose of the specification
to learn if you have violated the Standard.  [or found an error in
the specification...]

</blockquote>

I agree with Hixie that saying "specifications should assert a
tiebreaker rule" fails to rise to the level of good advice. It
doesn't belong in our quality guide. I have a long song and dance
about this but let me spare you the length of it at this late date in
the process.

I also agree with the QA WG that many groups do in fact have a
position on the authority and precedence of formal notation vs.
natural language exposition.  When such a precedence relation is
part of what the group believes about what they wrote, it should
be included in the governing writ.

Note the normative schemas in the Voice Browser products such as

http://www.w3.org/TR/2004/REC-speech-synthesis-20040907/

So what I am hoping for is a wording change that the Working Group
and the dissatisfied commentor can both find they can live with
[see above].

Al
/self (wearing no hat)
</discussion>

>
>--
>Ian Hickson               U+1047E                )\._.,--....,'``.    fL
>http://ln.hixie.ch/       U+263A                /,   _.. \   _\  ;`._ ,.
>Things that are impossible just take longer.   `._.-(,_..'--(,_..'`-.;.'

Received on Wednesday, 4 May 2005 20:17:56 UTC