[
Lists Home |
Date Index |
Thread Index
]
[Tim Bray]
>Certain people in the schema-enthusiast camp tend to condemn as "tag
soup" anything that hasn't been
>validated by a schema that is the result of 5 years of committee
deliberation, and I have little sympathy.
I think its important to remember that you can have tag soup that it
completely schema valid which
is probably worse than tag-soup in the sense of malformed tagging.
I say it is worse because the presence of a schema and the proud pronouncements
of "100% validity" can mask a multitude of tag soup problems.
This is what I was complaining about in this article: "The real cost of XML
tags":
http://www.itworld.com/nl/xml_prac/04042002/.
Schemas have entropy - a tendancy towards disorder - disorder the grows
over time
as the schema is modified. The disorder grows rather than increases becuase
modifications are made that loosen content models to accomodate new
requirements whilst guraranteeing that existing documents will remain valid
against the looser model.
I have seen much evidence in industry of schemas tending towards a collection
of foo :== (a|b|c...)* models as modifications are made to account for
changed requirements in the simplest possible way that is guaranteed
not to break existing documents.
The result of a tag soup. 100% XML, even 100% valid, but tag soup
nonetheless owing to the weakness
of the ordering constraints imposed by the content models.
Management wave their schema and congratulate themselves
on their XML smarts, little realising that with every self-important
exhalation, the key constraints on the ordering of data, that should
be in the schema, are migrating further and further into incomprehensible
imperative code and (gasp!) XSLT transformations in the bowels of the
business process from whence deep surgery will one day be
required to remove them.
Sean
http://seanmcgrath.blogspot.com
|
