[
Lists Home |
Date Index |
Thread Index
]
>> Russian doll schemas... A veritable dung heap. I am overpowered already :-)
> Hang on a second... what about Russian doll schemas makes them an
> xml-smell? I mean, I can understand situations where openness and shared
> structure are goldmines for interoperability, but I can just as quickly
> think of 10-20 reasons why you would want to localize definitions.
Ah, I was about to concede there and loosen my definition by saying
_when_ they smell.. but that does not make an interesting argument :)
So, I can think of reasons to localize element definitions. I can also
think of (legitimate) reasons to create a 30 method Java class instead
of breaking it up, but that doesn't mean I am 100% happy afterwards. It
works but it smells.
Of course a purist global complex type schema can be hard to stick to,
so in organisations where I have put in place schema design
guidelines, I usually put down the following:
- No global elements except candidate root elements
- No inline anonymous type definitions, all elements must reference a
global type except containers for repeating constructs because they are
so simple, e.g. businessCenters containing businessCenter*
regards,
Christian
p.s. To be fair, most of these schemas were on the large side (hundreds of
types or more). The consequence of the strict guidelines was good
reuse, the ability to create a glossary, no hidden information, and so
on... the design guidelines were enforced using an XSLT stylesheet
that created a quality report.
|
