[aggregate reply to multiple posts from the same person]
On Fri, Jun 5, 2009 at 6:38 PM, Steven M. Caesare<[email protected]> wrote:
> That's my point WoW ran Win16 code unaltered, which had no concept of
> privilege and permission. The VDM created for a Win16 session did a
> decent job at virtualizing the hardware, but file system access in the
> Win16 world required pretty much free reign.
Most Win16 applications didn't really care where they are in the
filesystem, precisely because the world of MS-DOS and Win16 was total
filesystem chaos. Some wanted to be in a particular application
directory (C:\FOO or whatever). None of that is really incompatible
with the idea of permissions.
Stuff that directly fiddled with the hardware was generally the
biggest problem, and much of that wasn't even Win16 but rather DOS --
DOS was so limited that banging on hardware was often the only way to
get anything done. Or stuff that expected to interface directly with
the Windows 3.x libraries and subsystems.
That kind of low-level stuff had almost as much trouble in the world
of Win95 as the world of NT 3.x. Almost all of it still required
rewriting. It wasn't just a matter of changing your makefile target
and recompiling. They ended up rewriting for Win32 on Win95. They
could have rewritten for Win32 on WinNT, but Microsoft was promoting
Win95 over WinNT.
Win95 had some real compatibility advantages over NT-as-released.
It had a built-in bridge to the world of DOS software that didn't like
Windows at all, i.e., "MS-DOS Mode". That shut down the Windows 4.0
stuff and reverted to MS-DOS 7.0 without Windows. But there's no
technical reason why NT couldn't have included a similar feature.
It's just an automatic dual boot.
More significantly, 95 could run even if you had hardware with only
real mode drivers. It would be slow, but it usually ran. That let
people keep using hardware that lacked new drivers. That's the sole
justification I can see for keeping the "Win Classic" line alive, as
far as compatibility is concerned. And it's a non-trivial reason, so
maybe it's even justified.
("Win Classic" is my name for all the Windows releases that depended
on DOS. It began as Win 1.0 and evolved to Win 3.x. Then it was
bundled with a copy of MS-DOS and called "Windows 95". It finally
died with Win ME.)
But utility and application software? That's a weak argument. Most
stuff didn't care. The stuff that did, didn't work right on Win95
either. Ever try running a DOS/Win16 defrag program on a Win95 box?
Heh heh heh...
> But clearly MS considered security important enough to build a pretty
> robust kernel infrastructure and file system to support it.
And then ignore it for almost a decade, right.
> Right or wrong, migrating that installed base is largely what allowed
> Windows to move in to the NT-based world.
Yah, I'm not seeing that. All the apps that didn't run right on NT
3.x didn't run right on WIn 2000 and XP. They were dragged kicking
and screaming on to NT -- some of them are still kicking and
screaming. If that process was started in 1995 instead of 2001, we
would have been where we are today years earlier.
Ironically, it's Vista that introduced real help in this department,
with filesystem and registry redirection. If only Microsoft had spent
the effort they put into Win95's internals into that, and made it part
of NT 4.0.
> But try to find any real user-centric
> Win32-based software around the time that NT came out.
NT introduced Win32; of *course* there wasn't an installed base.
But NT could run "well behaved" DOS and Win16 code, same as now.
> All the other software around was Win16 bases, and the stuff that was
> migrated to Win32 targeted the Win95-based platform which didn't
> have account privilege or a secure file system either.
You're putting the hose before the cart. They targeted Win95
because that was the OS Microsoft was offering as the best choice, and
the OS that was coming on all the new computers. Why the hell would
they target something Microsoft was positioning as a secondary product
(NT)? If Microsoft had focused on NT and released an improved NT 4.0
in the place of 95, the world would have targeted that instead.
> You must have some pretty straight laced devs that you know... ;-)
Not sure what you're trying to get at there...?
On Fri, Jun 5, 2009 at 6:51 PM, Steven M. Caesare<[email protected]> wrote:
> I'm not sure how familiar you are with the state of things in 1993, but
> Microsoft's dominance in introducing a new OS was not exactly a slam
> dunk.
I remember clearly when NT came out in 1993 and there was a question
of whether NT or OS/2 was the better migration path. Both could run
well-behaved DOS and Win16 code; neither tolerated direct hardware
access; both had stricter hardware requirements. It would have been
something of a toss-up.
I don't see Win 95 helping that much. There's still a compatibility
headache, just a different one. I don't see it making life for
Microsoft surer. Unless, perhaps, simply by mudding the waters with
an additional choice. But that could have gone either way: "With
OS/2, you only have one platform to target; with Microsoft Windows,
you have two."
I speculate (without hard evidence) that the continued existence of
Windows Classic (as Win 95) might have been the result of turf wars
inside Microsoft. I'd bet good money that the people responsible for
DOS and Windows Classic didn't want to give up their kingdom to the
new upstart, Windows NT. I also posit that turf war continued for
some time. How else can anyone explain Windows ME? HHOS.
-- Ben
~ Finally, powerful endpoint security that ISN'T a resource hog! ~
~ <http://www.sunbeltsoftware.com/Business/VIPRE-Enterprise/> ~
