Lists Home |
Date Index |
On Sun, 2002-01-27 at 18:30, Paul Prescod wrote:
> Mike Champion wrote:
> > I agree with Simon. For example, my canonical example of an order;
> > if you want to stay in business, you'll be liberal in the vocabulary
> > and structure you accept so long as it is well-formed and the
> > information you need to process and validate it (in the business
> > sense, not the XML sense) is in there somewhere.
> How will you write software to verify that the information you need is
> in there? Isn't the easiest way just to check it against the schema?
Easiest, yes. Only, no.
I'm constantly amazed by how different expectations for computerized
processing of invoices or orders are from expectations for human
Humans processing faxed or mailed documents are quite capable of
recognizing different fields in forms even when they arrive on different
stationery or different languages, and can even pick up the phone and
make an inquiry when there is a problem.
In some important ways, humans are far more adaptable than computers.
Programmers often seem to expect that there can be and will be only one
acceptable form for such information, and that their job is to map that
form to their structures and never think about it again.
I've hoped for a while that XML's flexibility might lead to the
development of approaches which balance or even combine human and
computer processing of information, combining humans' extremely flexible
interventions (aka "teaching") with computers' highly efficient
performance of repetitive tasks.
I can't say I've seen this notion taking hold on any large scale, though
every now and then I hear of encouraging bits. I don't believe this
problem is technical, however. It seems harshly cultural. Programmers
are deeply loathe to admit that users might have a greater role than
use, and that regular human intervention in the very logic of a data
structure might have advantages rather than disadvantages.
Maybe some day. I'm hoping that the Swing side of MOE will support some
of this kind of work, but I'm a ways off from my own implementation. In
the meantime, schemas (taken as deciders rather than advisors) feel to
me like a horribly wrong turn, combining all the virtues of hairshirts
Ring around the content, a pocket full of brackets
Errors, errors, all fall down!