[geeks] Prolog for network configuration
Joshua D Boyd
jdboyd at cs.millersville.edu
Thu Aug 8 13:38:20 CDT 2002
On Thu, Aug 08, 2002 at 06:18:08PM +0100, David Cantrell wrote:
> On Wed, Aug 07, 2002 at 10:27:30AM -0400, Joshua D Boyd wrote:
>
> > To further make that situation worse, DLLs need to use the same malloc
> > and free that the calling program does (which is not a simple thing to
> > pull off), or things are pretty much garanteed to go wrong.
>
> Is there not a single standard malloc/free in whatever the Doze equivalent
> of the standard C library is? If people are writing their own malloc/free,
> then I suggest that the problem lies with those developers and not with
> Windows.
It seems that there are a minimum of 2 Microsoft provided malloc free
sets. One for C and one used by new/delete (according to my memory of
the article, which was in Windows Developer Magazine, the new/delete set
were for some reason not built on top of the same malloc/free as the C
one was). And then further, the malloc/frees used by Borland C++,
mingw32, and cygwin are again all different. So, a DLL needs to find
out what the application is using and use it. Otherwise, things are
going to get seriously screwed up.
One could argue about the idiocy of the other compilers not just sharing
MSs malloc/free, but I think that MS makes it extremely difficult for
them to do so.
> > And of course, I've always liked that NT had a fairly fine grained ACL
> > system, for the FS at least.
>
> I remember reading that it's similar to VMS in that regard.
My understanding is that the ACL system applies to far more than just
the FS, and that it is all based on what VMS did. I've only been
exposed to the FS though, and have yet to touch VMS.
> Things went wrong when:
> they didn't have a console, and
> they put graphics drivers in the kernel
That second complaint didn't occur until version 4.0 BTW. Lack of
console was stupid. I think at one point NT might have had a console,
but it was never shipped.
> That Doze programmers have this nasty habit of writing stuff that "needs"
> to be run as root doesn't help either. Even worse is the amount of stuff
> that "needs" to run in kernel-space.
What needs to run in root and kernel space? I've never come across
anything but installers that wouldn't run under non-administrator
accounts. And I wasn't aware that they had added still more to kernel
space since the graphics moved there. That would explain some things
though.
--
Joshua D. Boyd
More information about the geeks
mailing list