I'm wearing my Register shirt as I type this so given I'm already apparently
a tool of The Register I might as well go all the way and
point you to an
article John Lettice there has written about Microsoft's seeming obsession with
Linux.
Personally I think Linux is crap. At least from a desktop point of view for
getting general office productivity done. But as a server, Microsoft needs to
worry. I think they have an uphill climb to win the argument that Windows is
less expensive than Linux due to TCO (total cost of ownership) because that is
only true while Linux specialists are hard to come by and hence can charge a
premium for knowing the arcane ways of Linux.
The real problem with Windows on the server is cost per license combined with
scalability. Microsoft has increasingly closed the door on third parties being
able to extend Windows.
To use a basketball analogy, it's like when you're playing a 4 on 4 game and
your team is losing. There is that temptation, when you think you're the best
player on the team, to try to do it all yourself. In the end though you'll lose.
I think there's an argument to be made that Microsoft is suffering a bit from
that. Rather than just open the door further and let third parties provide
solutions, Microsoft seems paranoid about security in the wrong places.
For instance, a lot of what makes Linux appealing for IT people is its
transparency. It's "simple" if not user friendly. Stardock's a Microsoft shop
and the difference between doing dynamic pages where you have X.wincustomize.com
on IIS versus what you can do on Linux out of the box is just one example of
where Microsoft envisions the OS being used in specific ways and tailoring the
OS for that. In contrast, Linux, because of its open development structure tends
to not have any particular focus which has the good side of making easier to
mold to a specific use.
Longhorn, from what we've seen early on, has us concerned because it may be
going even further in that direction. Microsoft recently made an announcement
saying that nearly half of all Windows crashes were due to non Microsoft code.
The inference, to me, was that if only people would stop using crappy third
party software Windows would be perfect. Of course, that's absurd. It ignores
that there's a lot more third party software being run than Microsoft and yet
Microsoft's code seems to cause the majority of the crashes. But worse,
it's the entire attitude that third party software developers are a pain for
them.
The security craze is really what's driving things right now over there
combined with a concern that Linux is making in roads. Put together, it's
creating a mindset, at least from here, in which third parties are really just a
potential security issue that they wish would go away. I could be wrong on that
but consider this -- Windows XP bundled zip, image management, CD burning,
instant messaging, movie making, skinning, and other software which cuts many
developers off at the knees. Longhorn is scheduled to come with a built in
database layer on top of NTFS, DVD burning, and virus scanning. And that's with
2 more years at least before release, who knows what they'll add in the
meantime.
But it gets worse than just the bundling. In the name of security, there are
talks of making much hard to extend the OS. Hooks and other ways third parties
use to extend the base feature set of the OS may be taken away to make the OS
secure (yea because that's where all those viruses come from! system hooks! Not
say Outlook Express email attachments and vulnerabilities in protocols...).
This won't have much affect on companies with the resources (i.e. money) to
spend the time to "do it right" (i.e. Stardock) but it'll close the door on
freeware and many shareware developers who just don't have the time or knowledge
to navigate the maze of security or digital signing licensing or whatever to
make innovative new stuff.
Consider this: If Instant Messaging hadn't already been invented and Longhorn
came out as currently planned you'd not likely see it because instant messaging
relies on a keyboard/mouse hook to determine when a user is at their machine.
Now Longhorn is likely to allow such hooks still because of instant messaging
but that's just the point, who knows what other things that won't be possible
because of this kind of "security".
And as a result, these developers are going to increasingly move to Linux.
You can see this already. I know of a few Fortune 100 companies who are moving
to Linux on the desktop as well as the server because Linux is easier for
them to customize.
Microsoft, in essence, is helping build Linux without realizing it.
So Microsoft fights back using the same tactics they used with OS/2. Except
those tactics won't work this time because Microsoft won primarily for 2
reasons: 1) They convinced third parties not to write for OS/2 -- that their
better opportunity was with Win32 and 2) They made it cheaper to bundle Windows.
You can't convince third parties to write to Win32 when Microsoft continually
cuts them down through bundling or removing their ability to do things with the
OS that Microsoft hadn't already thought of. And Windows can't be cheaper than
free.
So now they're moving to the total cost of ownership argument which only work
as long as Linux remains hard to use (relatively) and the number of people
familiar with it remains low. And Windows is no panacea. I still can't
figure out how to get Active Directory to work across domain controllers. 
If I were emperor of Microsoft I would be focusing on creating the APIs that
open the door to third parties to create the next generation software rather
than trying to have it all. I would also suggest trying to develop Windows such
that it works more generically. Don't try to envision how it's going to be used
all the time and make it difficult to use in ways that MS's team didn't think
of.
Just my 2 cents. BTW, I also don't think the security issues in Windows are
Microsoft's fault but I'll get to that in my next article.