[Bug 5164] validation vs assessment

http://www.w3.org/Bugs/Public/show_bug.cgi?id=5164





------- Comment #4 from johnarwe@us.ibm.com  2008-02-15 18:58 -------
I disagree with the resolution in comment 2.

wrt comment 1, I don't understand how the wg itself can be divided when the
spec itself (1.0 AND 1.1) defines the two terms to mean different things
explicitly in 2.1 Overview of XSDL (unless the wg asserts that "validation" is
not a derivative of "valid", which I don't buy either):

Schema-validity assessment has two aspects:
1 Determining local schema-validity, that is whether an element or attribute
information item satisfies the constraints embodied in the relevant components
of an XSDL schema;

2 Synthesizing an overall validation outcome for the item, combining local
schema-validity with the results of schema-validity assessments of its
descendants, if any, and adding appropriate augmentations to the infoset to
record this outcome.

Throughout this specification, [Definition:]  the word valid and its
derivatives are used to refer to clause 1 above, the determination of local
schema-validity.

Throughout this specification, [Definition:]   the word assessment is used to
refer to the overall process of local validation, schema-validity assessment
and infoset augmentation.

If you are going to define these terms, I think you are obligated to use them
consistently yourselves.  As a reader, it does not appear to me that they are
consistently used today.  If the wg believes they are, then it's fair to think
my reading is incorrect and during the next review I will have to carefully
think about the implications of this on its correctness...could be fun and
informative.

Editorially, it might be a simple partial fix to copy "schema validity is not a
binary predicate" from 5 to 2.1 where the existing definitions exist.

It is a separate, although very worthwhile, question to as to whether or not
schema users would benefit from having a common name for the most commonly used
assessment results, i.e. from the wg and spec defining a new term (or terms) to
capture the most common intent when "generic" users state that an instance
document is "valid" [wrt some schema].

e.g. [Definition:]   the word IYFNTH is used to refer to the condition where
all of the following are true:
- an instance document's content is assessed against a set of schema components
- the document's root element is the validation root
- the assessment invocation is type is element-driven validation 
- the validation root has a PSVI [validity] property value of 'valid'
- all descendants of the validation root have a PSVI [validity] value of
'valid' or 'unknown'
- the validation root has a PSVI [validation attempted] property value of
'full' 

I could see for example at least two useful definitions, one allowing for lax
wildcards to be missing schema components (above) and one requiring even lax
wildcards to have schema components (removing 'unknown').  My set of conditions
above might be an incorrect representation of this intent... if so, it just
makes my point by example.

If it is objectionable for some reason to put such definitions in Structures,
other venues like a Note would be acceptable to me.  I do see value in the
schema wg encouraging that kind of common understanding in practice, even if it
is not normative per se.

Received on Friday, 15 February 2008 18:58:40 UTC