begin  quoting Stewart Stremler as of Thu, Mar 24, 2005 at 12:27:57AM -0800:
> begin  quoting boblq as of Thu, Mar 24, 2005 at 12:06:14AM -0800:
> > On Thursday 24 March 2005 12:01 am, Stewart Stremler wrote:
> > > Shims, or thinking ahead in the design?
> > 
> > Surely you joke. Or are you omniscent? 
>  
> Not at all. Omniscence isn't required.

Let me emphasize something that I didn't hit very hard here -- I 
don't think that _everything_ can be foreseen.  Nor do I think that
trying to determine _all_ the problems ahead of time is a good idea.

Like most things, the extremes aren't the best solution.  Pure reaction
leads to unstable APIs, fragile code, needlessly complex systems, and a
lot of wasted effort.  Pure planning leads to inflexible APIs, unhandled
cases, needlessly complex systems, and a lot of wasted effort.

Every so often I've gotten it right.  More often I've been sucked into
the implement-it-now-think-about-it-later morass, or grown bored in the
let's-think-of-every-possible-case-without-seeing-what-really-happens
meetings.

There's a happy ground in there somewhere. Or so I believe.

A little thought, a little testing, a little development... all things
in balance.

-Stewart "Top-down problem decomposition/bottom-up tool implementation" Stremler
-- 
[email protected]
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Reply via email to