nemesis wrote:
> BTW, perhaps you'd be interested in this nutjob and his take on the 
> future of computers and programming:
> http://rebelscience.blogspot.com/2008/09/cosa-new-kind-of-programming-part-i.html 

I'd be much more interested in that if he'd actually written something 
larger than a quicksort. Or even a quicksort that would actually work in 
a generic sort of way, like with different types, or different size 
arrays, or whatever.

Too much handwaving while loudly proclaiming he's solved every problem.

He thinks hardware is very reliable, but he's clearly never worked at 
building hardware, where you ship off dozens of versions of chips, and 
when you're done, you use software to patch the bugs in the hardware 
design.  He also doesn't seem to realize that there are all kinds of 
timing problems in hardware, and that indeed solving the timing problems 
in "synchronous" hardware is a major part of the development effort.

He also thinks that hardware is more complex than software, but the 
reason people write software instead of building hardware is that much 
that software does is too complex to implement in hardware.

He also seems to think that the problems with making software work is 
problems with getting from detailed correct specs to software that 
implements those specs. Much more often, the specs on what a piece of 
software is to do is far more vague than what hardware is to do. Nobody 
starts building a chip before they decide what all the CPU instructions 
are going to be for sure. I'd also suspect it's pretty rare that 
hardware is designed in a way that's easy to upgrade - add-on floating 
point units would seem to be the most obvious example of what I mean, 
while every piece of software is meant to be "soft". Virtually nobody is 
happy with software that never changes once released, and for those who 
are (a la cell phones, microwave ovens, etc) it seems we manage a pretty 
good level of reliability.

He also seems to think that hardware doesn't have the same problems as 
software, yet http://www.opencores.org/ exists.

When he shows how you'd build an air traffic control system that's more 
reliable than what we have, or even a SQL RDBM server, using his system, 
I'll maybe start to worry about it. :-)



-- 
Darren New / San Diego, CA, USA (PST)

Post a reply to this message