Andrew Welch scripsit:
> One approach is: do nothing. Incur that cost if and when it's actually
> needed, don't waste your time/effort/money on it now.
I have never cared for this idea, and indeed there are many obvious
counterexamples. Tim Bray's stuff on internationalization (back
when it was a new and radical idea) showed that the cost of i18n at
development time is about 10%, whereas after the fact it's about 100%.
This is a specific case of a more general point: it is always simplest
to hard-code all assumptions, but it *will* force redesign further
down the line, and the more assumptions, the more redesign. Hiding
the assumptions behind an abstract interface doesn't help, for that
interface has hard-wired assumptions too. And so on.
--
John Cowan http://www.ccil.org/~cowan
Would your name perchance be surname Puppet, given name Sock?
--Rick Moen
