Lists Home |
Date Index |
On Tue, 30 Nov 2004 13:57:24 -0600, Bullard, Claude L (Len)
> Every few years, we toss out the old terms and attempt
> to reapply the technologies that didn't quite make it
> in the last cycle.
That's The Way It's Supposed To Be, I think. Code gets refactored --
toss out the stuff that really didn't work, try the technologies that
didn't quite make it again after a couple of iterations of Moore's
Law. Political ideas get co-opted (Barry Goldwater's speeches weren't
much different than Kerry's flag-waving one when he accepted the
nomination, Lyndon Johnson sounds like a Bush Republican in
You HAVE to change the names or people will expect backwards
compatibility -- bad ideas have to die, and sometimes good names are
part of the collateral damage.
Peter Hunsberger wrote:
> Sure, some hand waving as they invoke the magic term "SOA"
> and next thing you know your boss expects transparent data
> exchange with 500 new business partners to be up and running in a month...
There's no dispute that the the hype surrounding web services / SOA
has done a disservice. I think it's just part of our collective job
description to learn to separate out the hype from the reality and to
try to persuade those around us of what the reality is. It takes
awhile to get enough credibility to tell the Pointy-Haired Boss that
the hypemeisters are full of it and NOBODY is getting real-world
transparent data exchange with SOA, WS-* ... or REST and the Semantic
Web for that matter. Data exchange requires hard work, and at best
technology automates the grunt work, e.g. as XML removes the necessity
of defining YACC grammars and writing parsers for every random data
format need. Maybe semantic technologies will automate the process of
building or configuring transformers between diverse data formats, but
they will create a new type of grunt work - building ontologies that
define the mappings that automated transformation engines can exploit.
What's important is to keep an eye on the central ideas that have
persisted across Structured Programming, Object Orientation,
Distributed Objects, and now Service Orientation. Loose coupling and
data hiding are certainly two of these central ideas. Some ideas
sound good but don't work out, e.g. "location transparency" in the
distributed object world. Some ideas that seem trivial have the
immense advantage of being massively scalable (e.g. HTTP), and some
ideas that probably will never scale could turn out to be just the
ticket for limited domains (e.g. the semantic web). The thing we can
do is to exploit the ideas that actually work, irrespective of whether
they are fashionable or not; and to ignore those that don't work,
irrespective of whether they are tangled up in standards or products
with others that do work.