> >The question I'd like to pose to the newsgroup would be what makes
> >Windows so unreliable and prone to crashing?
> Awful, isn't it? I was using operating systems 25 years ago that
> were more stable and better performers than Windows. It's not like
> Microsoft was blazing a new trail and had no prior art to draw upon.
My complaint against Microsoft is that anyone presuming to write a new operating system (or any other
program for that matter), unless doing it as a student exercise, should make it better than what had gone
before. Thus, someone in the field of language design once said that Algol-60 was years ahead of most of
its successors. Likewise, the Burroughs B5500 series computers were years ahead of most of the computers
that came after them. Perhaps the same can be said about UNIX and now Linux. Certainly, Windows is about
15 years behind the time.
I have been running Unix and Linux systems since the early 1970s. At the beginning, Unix was pretty
unreliable and would crash almost every day. By 1980, crashes were unusual. I worked in a department that
supported a bunch of PDB11/70s and VAX-11/780s. We had a large community of users (perhaps a hundred or
so) doing software development both for those machines and cross-compiling for communication systems such
as PBXs and stuff and about a dozen machines. We had, perhaps, a crash a week (total, not per machine).
By 1990, I was running a Dell box with a version of Unix provided by Dell. I think they wrote it. It
worked 24/7 for the four years I worked at that company and the only times we rebooted it were to change
hardware (we were doing a communication system and we needed to add telephone interface boards (Dialogic,
not modems) and memory once in a while). The software never crashed at all in 4 years.
Quote:> One of the major problems with Windows is that the OS is not insulated well
> from applications. This is one of the cardinal rules of OS design, a user
> application should not be able to corrupt the operating system.
As Fred Brooks and David Parnas have been pointing out for about 3 decades now, separation of concerns
and information hiding key components in the management of complexity. And management of complexity is
the major stumbling block in designing good systems, and has been since, I would guess, about 1975.
Microsoft has made no contributions to management of complexity.
Quote:> Notice how
> on Linux systems, all you usually have to do to install an application is
> plop the files into a convenient directory and execute. However, on a
> Windows system when apps are installed they almost always modify the OS
> by replacing critical system DLLs, making entries in the registry, etc.
The fact that Windows can manage only one .dll of a given name is a killer. When I first got my PC with
Windows 95, I bought Microsoft Office Professional and Microsoft Visual C++. I could not run both.
Whichever one I installed last would work, but they had conflicting .dlls. The goddamned company's
products were incompatible with one another!
When I tried to get support, I had to talk to either an Office type or a Visual C++ type. Each one told
me which .dll to load. When I pointed out that I had to use both products, they could do nothing. There
was no one who had a larger view there. Incompetant company. I solved it by switching to Linux, of
course, where I could have had both versions of the .dlls (.so's in the Linux and Unix case, of course).
They do not even do dynamically linked libraries correctly. I assume they never will.
--
Jean-David Beyer .~.
Shrewsbury, New Jersey /V\
Registered Linux User 85642. /( )\
Registered Machine 73926. ^^-^^