Re[2]: [Am-info] O/S 2 support - slightly off topic (my apologies)

[Eric M. Hopper]

On Sat, 2002-03-16 at 12:52, Gene Gaines wrote:
> Wow, an informed discussion with "Windows" and programming
> design/structure in the same sentence.
> 
> I was getting resigned to that being a thing of the past.

I'm giving up though.  :-)

While it's likely true that the OS implementation of multithreading is
better than Windows 95 or Windows NT's, the point is that multithreading
is a bad idea in the first place.

A bad OS implementation of multithreading doesn't cause instability
unless it's _really_ bad.  I don't think it's that bad in NT, or a
heavily loaded NT server wouldn't last more than 5 minutes.  As it is,
most of them can last at least an hour or two.

The thing that causes program crashes and hangs with multithreading is
bad application usage, which is amazingly hard to prevent, and requires
programmers of exceptional talent and skill.

So, as an OS innovation goes, multithreading isn't as useful as most
people think.  In fact, in almost all cases, you can do a better job of
making your application snappier and more responsive without
multithreading and its attendant tendency to make your application
unstable and unpredictable, even in an SMP environment.

I find Windows NT reliance on multithreading for any kind of performance
at all to be indicative of poor OS design.  In fact, a lot of their
low-level OS design seemed to be there to cash in on computer science
fads instead of to be good.

I just tend to get irritated when people wax nostalgic over OS/2,
because in my experience, it wasn't very good.  Better than Windows,
yes, but that's not hard.

Have fun (if at all possible),
-- 
The best we can hope for concerning the people at large is that they
be properly armed.  -- Alexander Hamilton
-- Eric Hopper (hopper@omnifarious.org 
http://www.omnifarious.org/~hopper) --