xmlhack: RELAX NG gets "feasible validation" support

James Clark has announced the addition to his Jing validator for RELAX NG of support for "feasible validation" (a type of "gradual" or "progressive" validation), as discussed in Rick Jelliffe's When Well-Formed is too much and Validity is too little presentation at XML Europe 2002.

Progressive validation support is useful if you have documents that are still incomplete or "under construction" but want to check that you've not made any validity mistakes in them so far. In the announcement, Clark states the option will cause Jing to:

first transform the schema by wrapping each element, attribute, data and list element in an optional element and then validate against this transformed schema. The net result is to check whether the document could be made valid by inserting additional elements and attributes.

In a comment posted to earlier xmlhack story about fuzzy validation, Rick Jelliffe elaborates by describing some scenarios in which you might find "feasible validity" checking useful:

You may be marking up some text and intend to add the "required" metadata later: you just want to find out if there are problems (elements names wrong, etc) in the elements you have done. You don't want to be bored by "errors" which are really due to your workplan, nor (worse) have the validation fail and stop before it gets to the elements you have been working on.

In the abstract to the paper Jelliffe presented at XML Europe 2002, he outlines three techniques for checking the validity and well-formedness of documents that are incomplete or still under construction:

weak validation

A technique based on "strength-reducing schema (or DTD) particles so that all elements are optional. Again, only infeasible documents will cause validation errors." It will thus raise errors if something forbidden by the schema is included, but allow mandatory elements to be missing. Clark's new option in Jing implements support for weak validation.

partial ordering

A technique based on "finding which tags (start- or end-) can feasibly appear before or after each other; unfeasible markup can be discovered before the document is even well-formed." Partial ordering (which can be expressed very simply using Jelliffe's one-element Hook validation language, described as "like a checksum for a schema") checks for a more limited set of contraints; it simply attempts to answer the question:

Does this element have a feasible name, ancestry, previous-siblings and contents?

Partial ordering validation is thus useful as a means for checking for errors even relatively early in the process of creating/editing valid documents.

Schematron phases

A technique which provides "a managed way to express many kinds of different constraints, allowing documents to be validated first against some criteria then others, suitable to the document's progress through a markup process."

Jelliffe contrasts all three of these validation techniques -- which, as a class, might generally be called "gradual" or "progressive" validation techniques -- with the techniques of checking for all-or-nothing strict validity and/or "partial validity" (which requires that "all the elements in every element must match unambiguously part of the way through the content model of the element") and states that "the first wave of XML editors [that is, most existing XML editors] required well-formedness and usually provided partial validity only, or even required strict validity."

Topologi, a company Jelliffe founded, implements progressive validitation in two applications: the Topologi Collaborative Markup Editor and the Topologi Schematron Validator (free download), which, while it doesn't check for feasible validity per se, does support Schematron phases.

Related stories