On Tue, 29 Mar 2005 11:37:55 +0100, ThePicMan <thepicman@infinito.it> wrote:
> 2) We mean two different things with "Cooperative Memory" here. I simply
>    mean that virtual memory must be an option for an application, and not
>    the rule. Virtual memory has promoted abusive allocation/use of memory,
>    making the systems MUCH slower than they ought to be. And all of this
>    to have, on average, just twice the addressable readable/writable memory
>    space (usually one doesn't have a 200GB swap file on a 512MB RAM system,
>    also because that's simply not really useful on 32bit ~Intel CPUs even
>    when using 36bit paging.. However, I see a lot of use for virtual memory
>    (I am thinking of file mapping) for specialized scientifical applications,
>    for example. But they must be designed properly). When I ask for the right
>    to allocate *physical* memory, I also want a way to say to the OS that I
>    don't need to access it for a certain timeframe (so that it may swap it
>    to disk, if necessary). This is one of the few things that were normal to
>    do on the 16bit Windows, which was cooperative in a lot of ways. Now we
>    have applications that think that memory is unlimited, and they make the
>    worst possible use of it.
>    It's simply ridicolous that a WindowsXP system is much faster (read:
>    less slow) when you add a 128MB stick to your 128MB PC. RAM is RAM, CPU
>    is CPU. CPU is slowed down terribly by virtual memory as-we-know-it. It
>    is a non-sense.

That reminds me of my use of MacOS up to about System 7.3 or so. 
Virtual memory was off by default and I never turned it on on any of
the machines I had because it slowed everything down so much.  You
could set the minimum and maximum RAM usage for each application and I
would tailor that to my needs.  That careful use of the Mac is still
with me today when I use a PC.  I am fairly careful to avoid opening
and closing applications out of order to avoid memory fragmentation. 
That was more important in Windows 95 than it is today with improved
virtual memory handlers.  It would, at times, be nice to have more
control over that functionality, though.

As to stability, I recall setting up a relative's computer to use a
ram disk for the Netscape Navigator cache because it would crash so
often and always damage the file system.  Once I did that, there was
no appreciable difference in the experience, but the crashes wouldn't
hose the file system.  That was a Performa series, IIRC.

Bradley
-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist