Re: [feyerabend-project] some thoughts
- Dave wrote:
> At the risk of boring others to tears....You are not boring me. The co-evolution of machines and languages has found an
unsatisfying local maxima. We will find a better way out if we consider both
Computers have been so optimized for the general case that any uniform handling
of special cases is judged too expensive. Take the processor's register file for
example. If we know ahead of time what instructions will execute then we can
allocate registers ahead of time too. This works great when the general case,
the normal flow, is all we ever need. So how many months do we spend a year
working out that general case? How many more do we devote to dealing with one in
a million cases that are now ten times harder to handle? How much end user
productivity is lost to the general case not being fast enough? How much is lost
to failures on special cases? I can only guess. I'd guess we've worked on the
general case way too long.
v 503-245-5633 mailto:ward@...
f 503-246-5587 http://c2.com/