Lists Home |
Date Index |
The sad fact of the matter is that the majority of the software development
community suffer from premature optimization tendencies. Its like an
illness we fight all our lives. I'm still fighting it but at least now,
after twenty years, I realize I'm ill :-)
Vendors know all about this predisposition to myopic optimization, so they
give the market what the market wants (as opposed to what the market
actually needs) which is stuff that feeds the urge to optimize. Phrases
like "tight integration" or "industrial strength" are used to deflect
attention from what is really going on - more often than not - dubious
engineering. It amounts to the same thing - feed the craving for
optimization - however ill conceived it might be.
Binary XML and 95% of the whole Web Services ballyhoo are classic examples
in my opinion. Who wouldn't be interested in an X-fold increase in
performance? Surely you get that by proactively micro-managing each aspect
of the system for speed? XML parsing? Yup, thats a definite perfornace dog,
lets optimize the hind legs off that particular dog.
I have long held out hope that the big lessons of HTTP would be
(a) keep it readable (design time)
(b) maximise statelessness (design time)
(c) scale it horizontally (deploy time)
I'm on the verge of concluding that this message is a stillborn. A real
tragedy given the amazing scaleability of the Web. If ever there way an
example of how to "optimize" an IT system the Web is it :-)
So why doesn't the message get through?
Perhaps because it is not obvious where the $$$ are in a "keep it simple,
keep it readable, scale it horizontally on the cheap." view if optimization.