Re: Markup for testable assertions

It would be *really* cool if we can do some machine processable assertions.
We can probably get some ideas from JUnit/CUnit stuff .. is anyone
interested? In any case, if we're going to put annotations in the XSD to
clarify all the actual constraints that go beyond XSD, then we have to come
up with XML syntax for the annotations anyway. Assuming that that wasn't
<annotation>English text</ann> then we need to figure it out anyway.

Is there a precedent for machine processable assertions in W3C specs that
you guys are aware of? I'm sure there must be a language someone has come
up with that we can mostly re-use.

Sanjiva.


"Jean-Jacques Moreau" <jean-jacques.moreau@crf.canon.fr> on 03/28/2003
03:22:53 PM

To:    "Amelia A. Lewis" <alewis@tibco.com>
cc:    Martin Gudgin <mgudgin@microsoft.com>,
       jeffsch@windows.microsoft.com, roberto.chinnici@sun.com, Sanjiva
       Weerawarana/Watson/IBM@IBMUS, www-archive@w3.org
Subject:    Re: Markup for testable assertions



+1 to Amy's approach. However, I'd go for a simplified markup; the
proposed one is just too cumbersome to type.

What about instead:
   <must>...</must>
   <should>...</should>
   <may>...</may>
   <mustn>...</mustn>
   <shouldn>...</shouldn>
   <mayn>...</mayn>
?

JJ

Amelia A. Lewis wrote:
> Can the xmlspec DTD be enhanced, either experimentally or locally to
> WSD, to include a <testable> or <assertion> element?
>
> This would, of course, also require a modification of the xmlspec.xsl
> stylesheet to handle the assertions.
>
> My preference would be that the testable assertions appear in the
> document itself, and that they be marked as testable assertions.  I
> would then like to see the stylesheet automatically generate an appendix
> on conformance, which would extract the 'tags' (an email message MUST
> have lines of no more than 998 characters plus CR and LF) and generate a
> hyperlink to the assertion in context.
>
> More below ...
>
> On Wed, 26 Mar 2003 09:29:40 -0800
> "Martin Gudgin" <mgudgin@microsoft.com> wrote:
>
>
>>This mail is intended to start a discussion about testable assertions
>>and associated markup in our spec. Here are some thoughts/questions:
>>
>>1.  Would it be better to have a section in the spec with all the
>>assertions in. And reference those assertions from their 'location' in
>>the spec itself? Or would it be better to 'sprinkle' the assertions
>>throughout the spec?
>
>
> Sprinkle.  Consolidate in appendix.
>
>
>>2.  Do we want the assertions to appear in the spec itself or is
>>there a separate stylesheet which emits the assertions?
>
>
> Assertions SHOULD be part of the normative text.  The stylesheet SHOULD
> generate an appendix which consolidates all of the assertions into one
> easily referenced section.
>
>
>>3.  Do we want 'classes' of assertion? Seems like whereever we have
>>things like MUST/SHOULD/MAY then we have an assertion. Seems also we
>>would want to capture the distinction in the markup.
>
>
> Seems like a good idea.
>
>
>>4.  Some assertions are captured in the schema. For example the fact
>>that wsdl:import and wsdl:include must appear before wsdl:types
>
>
> This is also in the text, is it not?  In fact, the text is far clearer
> on the subject of required sequence, I believe.
>
>
>>5.  Some assertions are captured in the schema for the 'single WSDL'
>>case but not in the 'multiple WSDL' case. For example, the uniqueness
>>constraint on the local name of port types is enforced by the schema,
>>but in the face of wsdl:include you could end up with a collision,
>>which would be an error.
>
>
> But the assertion appears in normative text as well, does it not?
>
> Amy!

Received on Friday, 28 March 2003 06:22:41 UTC