RE: [xml-dev] [ Conference ] What is Good XML? How Can We Tell?

["Len Bullard" <cbullard@hiwaay.net>]

I find myself drifting back to this one given the current job.

QA is a human concept because quality is.  Unless one reckons with the
humans-in-the-loop for a given XML production, most of the interesting
answers will be computer-sciency answers (we know all the geek words) to
questions of relatively little importance in the sense that they are
answered by the design of XML itself.

Some short answers in hopes of provoking a flame war:  :)

1.  "XML systems and software are complex and constantly changing."

I would say XML applications realized in software as systems are.  XML has
changed very little.  That's why it keeps working.  The Latin of Geeks.

2. "XML documents are highly varied, may be large or small, and often have
complex life-cycles."

XML document features in theory make this easy to manage where the only
output of a lifecycle at some state is well-formed first, but since that is
sine qua non, lifecycles are measured as validity or conformance to a
definition of some type.  The type determines strength of the control. E.g.,
XML conformance to a DTD vs a Schema vs text.

Quality is not a measure.  It is a requirement of a measure.  How do
measures relate to quality requirements? 

Let's say a given cycle contains edit/save/transform/load.  A fairly simple
cycle but a common sequence (a separate topic is do you want to measure
quality as node sequences, that is, a measure of a cycle length in
relationship to other cycles).  What is being tested at the node
transitions?  What happens in the / to make the adjunct nodes stronger or
weaker?

XSL both transforms and stripes (retags and adds literals.  How does XML add
quality?  Validation (well-formedness assumed).  Assuming the relationships
of the node validity requirements are tuned.  Here quality is defined by
negotiation.  What does XML contribute to the quality of the negotiation?

What about the text?  A much understudied part of the quality topic.

On the other hand, most quality definitions for text are in part, copied
from other sources.  How does XML contribute to a measure of quality for
some measurable in this life/cycle?

Same ones:  well-formedness and validity.  


"       . Quality in documents, document models, "

Mostly the computer sciencey answers.


"software, transformations, or queries"

XML as applied/practiced.   The challenge I see in practice is the quality
of the social/professional networks that create and arbitrate.  Negotiation
as a human practice particularly given tool choices.  

Things do break here.  Why?  The social aspect of the human user that
inclines one to favor local processes and practices over network-wide cycle
measures.   They attempt to do the least work possible.  So at a cycle
exchange, they measure quality backwards to source instead of as a
contribution to the next cycle.  In other words, what is quality for the
tech writer using Word and the contributions the Word sources can make to a
preconfigured set of XML documents, transforms, entities if measured
backwards is suspect.  

Spend an afternoon explaining to a QA specialist trained to measure
backwards and does not understand the XML systems definitions (speakGeek?)
in place by contract why they can't have that figure where they want it.
Aside:  there is a cost for content tagging.  This is one argument where
that cost is hard to reckon but easy to see:  the tagger is beaten.

"        . Theoretical or practical approaches to measuring quality in XML

Quality can be recognized.  But to measure quality, you must define the
measure.  An abstract measure is
backward-coherence/forward-contribution-to-next-cycle if the objective is to
measure affect over network coherence or sustainability.


". Does the presence of XML, XML schemas, and XML tools make quality 
"  checking easier, harder, or even different from other computing " "
"environments

Yes.  For it to actually work well, the XML specialist has to actually know
what they are doing.  Saying you have XML skills because you can edit in an
XML editor and render to PDF is saying you are an architect because you have
an axe.   To render quality in performance and production of XML, one must
be a smarter dog.  Smart dogs cost.  Quality is never free.  It is sometimes
friendly.   Do not Do not Do not underestimate the chances of hiring good
XML help if you piss them off.

"        . Should XML transforms and schemas be QAed as software? 

Yes.  Particularly because of contributions.

"Or configuration files? Or documents? Does it matter?

Nope. Pick something that works although I think of configuration files as
XML documents.  I think of XSL as a functional description.

All that said:  the customer measures are critical to understanding XML.
This is where all the social network aspects as contributors or detractors
of quality burn most hotly but XML can render almost meaningless as long as
the measures on both sides are crystal clear and acknowledged.  And here
discussions of quality get to the human root:  if the humans don't care,
nothing the machine can tell them will improve the quality of the product.

Thanks for reading this far.  Noodling on the thrash.

len