Re: [Chicken-users] inventory of community skills

[Shawn Rutledge]

On 2/5/07, Brandon J. Van Every <address@hidden> wrote:
> Shawn Rutledge wrote:
> >  And it needs to run at
> > maximum efficiency at every scale of device.
>
> Why?  Microsoft owns most of the world and doesn't share your passion
> here.  I've gone bankrupt on premature optimization.  I'm wondering
> where you think you'll get the support resources for such a lofty goal?

Least of all do I care what MS thinks because I don't have any passion
or admiration for most of their stuff; do you?  Their "ownership of
the world" is mostly a matter of critical mass and entrenched habits
on all sides.  Anyway as you know "maximum efficiency" is an
asymptote, a matter of continuous improvement.  Most big-time software
organizations are not really bothering much with that; they just bury
the details under ever-thicker layers of abstraction and try to forget
about them, and if the low-level stuff is running too slow, who cares
because machines keep getting faster.  But their software gets bigger
and slower at a faster rate.

I think possible definitions of "premature optimization" are
optimizations that later become obsolete, or that require more time
from you to optimize the code than they will save you in cumulative
runtime, or that bog you down for too long and prevent you from
writing the more useful parts of the code.  But sometimes you can
predict which parts will not be obsolete anytime soon, so doing it
right the first time (when you have the context in your head and know
exactly what you want, as opposed to later when you've forgotten and
moved on to higher-level things) pays off forever after.  Another
thing is that some decisions can be mislabled as "optimizations" when
in fact they are design features, and putting off the "optimization"
will require rewriting too much.  Picking the right algorithms,
libraries, number system, appropriate model designs, etc. tend to be
like that.