Strengthen Exit Criteria

I have some concerns about the exit criteria for XForms 1.0 listed in the
section of the document titled 'Status of this Document'. Of specific
concern are the two statements:
1. Sufficient reports of implementation experience have been gathered to
demonstrate that XForms processors based on the specification are
implementable and have compatible behavior.
2. An implementation report shows that there is at least one implementation
of each feature.

My basic concerns with these statements are...

Statement 1 needs more definition than 'sufficient'.

Statement 2 falls way short of what will be needed to ensure that XForms is
viable and usable. Interoperability is required to establish that a feature
is not only implemented correctly, but specified correctly. Also, to have an
implementation of each feature does not establish that the feature will work
in real world situations or will make sense in the context of an
application.

Changes Needed:

I believe the following changes to the exit criteria are needed to address
these issues.

1. There must be a test suite that specifies a test for each feature of
XForms plus many interaction features. See
http://www.w3.org/Graphics/SVG/Test/BE-ImpStatus-20010127 for an example of
exit criteria that specifies this.

2. There must be more than one implementation of each feature and there must
be interoperability between each.

For an example of this, take a look at the CSS3 Selectors test suite at
http://www.w3.org/Style/CSS/Test/CSS3/Selectors/20020115/. "There must be at
least two interoperable implementations for every feature".

Note the definition of 'implementation' in this test suite which includes
'available', 'shipping' and 'not experimental'. I think it is a good one.

3. There must be real application of the spec. This goes beyond building a
designer tool that generates XForm markup and a viewing tool that displays
it and allows a user to fill out a form and submits it. There must be people
who implement real projects using tools that generate XForm markup and tools
that allow users to fill out forms and process the results on a server.
There needs to be people building credit card applications, order forms,
surveys, and all kinds of applications. Each of these has the potential to
uncover new problems with the specification that were not apparent in the
design or even the implementation of the tools.

The Canonical XML recommendation (http://www.w3.org/TR/xml-exc-c14n)
specifies a stronger implementation criteria as in 2 above and satisfaction
in an application context as this item 3 insists. From Canonical XML: "at
least two interoperable implementations over every feature, one
implementation of all features, and one report of satisfaction in an
application context"

It is a goal of XForms 1.0 to make forms declarative such that many forms
will not require scripting. This will not be remotely achievable if each
person publishing a form has to plug in script to work around the holes that
were not exposed by this kind of real-world testing. We could end up in a
situation much like we have today with HTML forms where every form requires
script. This will significantly slow the adoption of XForms into the areas
it needs to go or halt it's adoption completely. That would be a colossal
waste of everyone's time invested in this specification.


A Call to Action:

A natural objection to these comments is that it will take more time. It
will. And we do have to be concerned about taking too much time. But
somewhere we have to find a balance. And for some of us, this is simply a
matter of will. Are we willing not just to finish an XForms specification,
but to do what is necessary to see it successfully adopted?

Considering the great benefits to the entire community of achieving
recommendation status sooner rather than later, Cardiff is prepared to
contribute to the development of a test suite and to move aggressively to
implement the design side of an implementation.

Who else will contribute to the development of a test suite?
Who will commit to developing viewing tools?
Who will commit to using these design and viewing tools to build real world
applications?

I am convinced that the best thing we can do for XForms right now is to all
move aggressively to invest in all of these areas. All of these things are
needed for XForms to ultimately be successful.


J Joel Faul			Director, Internet Product Development
jfaul@cardiff.com		http://cardiff.com

Received on Friday, 22 February 2002 13:57:01 UTC