🚀 go-pugleaf

tonyb@titania.juliet.ll.mit.edu ( Tony Berke) writes:

>I've noticed what appear to be follow-ups on this issue, but not the
>answer to the basic question: Can I only write Windows programs that
>will run on 386's ??   If that's true, I assume it has to do with
>Windows' bogus memory management scheme, and Borland's desire to make
>it a non-issue.  [...]

Don't assume that everything you don't understand is bogus.

Windows memory management is built in a way which allowes moving
around memory blocks behind the back of the program. This is a 
usefull and necessary feature. Its implentation is based upon the
segmented architecture of the processor. Unfortunately, it requires
some cooperation from the programmer (proper locking and unlocking
of memory segments), because the 8086/88 processors implementation
of segmentation is too simplistic: segments are adresses, not names
(indizes) of addresses, which makes it impossible to move around something
a far pointer (a segment/offset pair) points to, without changing
the actual segment register content. On the 80286 (and above) in
protected mode, this is different, now segment registers point to
descriptors, which in turn point to memory. This makes it possible
to reorganzize memory without the programs cooperation. The nice 
think is that no code changes are necessary for a Windows application
to run in protected mode. Only the necessity for proper memory locking/
unlocking has gone, if you don't care whether your program still
runs in real mode. Obviously, the whole memory management has been
designed for the protected mode of the 80286 up processors, which
required quite a lot of foresight, in my opinion.

Wolfgang Strobl
#include <std.disclaimer.hpp>