[
Lists Home |
Date Index |
Thread Index
]
Ronald Bourret wrote:
> Yes. I've seen it in at least one other organization as well. My point
> is that it hasn't crossed schema boundaries and become universal in the
> way people thought it might. (There might very well be a good reason for
> this. For example, given the potential complexity of addresses, somebody
> designing for a local market might be making a very good design decision
> to ignore all that complexity and simply encode the address schema that
> fits their locale.)
I see two issues to consider beyond the usual interop concerns;
dependency management and engineering cost.
Mapping XML content models can represent significant work, yet it's
often desirable to reduce outside dependencies. Many groups don't
reuse schemata because they're wary of being broken by another
spec outside their control. For example, the Atom effort has a large
percentage of elements that could be taken from other specs such as
dublin core - the consensus nonetheless has been to retain control
of the spec through re-invention.
Normally the focus is solely on interop, but it's a mistake to
ignore the costs of supporting generic formats. When you do decide
to reuse, some uber-content standards* can be so generic and are
trying to cover off so much ground you risk overspend and system
robustness in simply being conformant. Which is to say the
interop/implementation costs can be high enough to constitute
overengineering and can put systems and projects at risk. The agile
folks call this design speculation "speculative generality". The
ideal approach seems to be profile for the target locale, which may
imply some level of governance or architectural support.
So, it's not just that reuse and interop are good, but that there
are dependency risks to consider plus the engineering cost of all
those SHOULDs and MAYs add up. This is why for example,
architectural policy in Propylon has always been to make
transformation cheap as possible rather than hold unjustified
expectations about standard models and format reuse.
Having said that, where I am seeing reuse working in is the Irish
eGov scenario Sean is involved with (some of the RIGS have popped up
here recently). The essence there is to to profile existing
standards and ensure the architecture supports those who would use
standards. Sean might have a more nuanced view on this, but it seems
to me without astute profiling, good IT governance, and
architectural support for standards, format/schema reuse is fraught.
cheers
Bill
* such as xAL, WXS (or even ISO8601)
|
