Roger makes me look like a total wimp by posting _this_ of all things to
the xml-dev list:
> In fact, the design of data
> to be exchanged matters very little, as long as all the data that is
> needed exists. At the consuming end software can extract the data and
> store it in memory in a form suitable for efficient processing.
> Software is king! Data design is only needed to the extent that you
> document and define what it contains; other than that, do whatever
> you want. Software design is important, data design is not.
Practically every programming list in the world would eat that up and
send thank-you notes, but here, not so much.
One of the fun things I've encountered watching the Erlang folks is that
even though the language spends a lot of time making binary data formats
(and ASN.1) easier to work with, changes to binary data formats are even
less popular than new versions of XML formats. It's a lot more work to
make even the basic changes and make sure they're correct.
In a sense, by switching to text, markup gives Roger's argument more
power, as it lowered the switching costs. Transformations are
relatively easy and widely supported, and the history of SAX processing
in particular suggests that decoupling external formats from internal
structures isn't that hard or that unusual.
I still think data design is a drastically undervalued field, one that
needs a chance to develop virtues of its own.
Thanks,
--
Simon St.Laurent
http://simonstl.com/
