[Microsoft's New Win-Win Strategy: Post 3 of 5]
My last posting, "Scrap The Windows Codebase" explained some of the reasons why the current Windows codebase might be better off in the trash bin. My mention of "Linux" caused myriads of knee-jerk reactions from other bloggers (see the Channel 9 forums, for example). But, this post will only talk a little bit about Linux, and a lot about what might be needed for the future, and why the operating system has become too central a focus for Microsoft.
What IS Microsoft's Business?
I'm not sure Microsoft ever wanted to be in the operating system business. There is been a lot of revisionist Microsoft history floating around, and the original goals of Gates, Allen, and eventually Microsoft are sometimes lost in the rhetoric of their successful business strategies.
Today, Microsoft knows their operating system is the cement that glues their business strategy together.
Microsoft executives describe how desktop applications "widen the 'moat' that protects the operating system business". At the same time, Joachim Kempin in his 1999 testimony said "We are not in the operating system business. We are in the computing business." While at first glance this seems like defensive rhetoric to distract Jackson's team from their OS focus, keep in mind that Kempin was with Microsoft since 1983, and his chief responsibility was developing their relationships with PC manufacturers and OEM distributors. Kempin, perhaps more than anyone else, saw the operating system as a useful business tool, one which wielded legendary power over extending Microsoft's computing products to every desktop.
While there is no doubt that this strategy has yielded tremendous financial rewards for Microsoft, it is nonetheless as strategy based, not upon technologies or innovation, but upon tie-ins, bundling deals, and partnerships. In this sense, it limits Microsoft because, unlike other companies, they do not need to create the best products, only the most viable ones suitable for their (so far) successful business strategies.
But, when Gates is at his best talking about Microsoft's strategies, there is not a mention of predatory business practices, and the "moat" that protects the operating system. Gates, in a notable 2003 interview with Fortune, said t "One was our vision, which has not changed since the day the company started." The vision was, according to the interview, "the idea that you could buy PCs from many different hardware companies, and yet they would all run the same software".
C'mon Bill! Let's look back briefly. Bill originally thought computer languages were the company's business. He went to IBM for an Operating System. Only when IBM couldn't come to terms with Digital Research for an OS license, did Gates see an opportunity, buy QDOS, and license it back to IBM. This sounds a bit closer to what Kempin might describe as "the vision": bundling and aggressive licensing.
All in all, I still believe Gates wants to serve the needs of consumers. One of the advantages Apple has over Microsoft is that their vision of ease-of-use and style is strong by comparison to the muddled vision of Microsoft. This wavering statement of vision, and inability to reconcile what Microsoft says with what it does is one of the reasons that so many Microsoft-bashers conclude that Microsoft is simply greedy and predatory. Without clear vision, all successful business might appear so.
At his best, Gates talks of the dream "of a PC on every desk and in every home". I think he still believes that comptuers are good for people and that the business mission is to do whatever is necessary to enhance the computer experience.
What does all this have to do with operating systems and my Windows TNG idea? Well, two things.
First, I believe Microsoft has become so distracted by the importance of their operating system as the glue for their predatory business practices and the "moat" that they have stopped innovating. Worse, the size and complexity of the operating system itself has slowed down even progress on Microsoft's business strategies. If ever Microsoft needs innovation, now is the time.
The second point Bill makes better than I can. In the Fortune interview, when asked "What can Microsoft do for small business?". His answer, in full, is
Making our software simpler will probably have more dramatic impact with small business than anywhere.
Anyone who is following the current Vista releases knows that Microsoft is not moving toward simpler, but toward a far more complex, multi-faceted operating system. The OS has taken over the company, and it's taken over the consumer's view of the product line. Windows TNG is one perspective on how Microsoft might change that.
A New Platform
If Microsoft is going to return to the goal of extending the power of the PC and creating greater consumer value, they need a new platform. What might that platform look like?
For a moment, let's speculate about the feel of a new product line, and the technical underpinnings. Consider the profile of the products and technologies I'm suggesting, and don't get hung up dwelling on a particular flaw or inconsistency, as there will be many This is, essentially, a "White Paper".
It's first worth considering exactly what the criteria might be for a new Windows.
Goal 1: Unbundling
Windows may not be too big if you consider all that it does. And Windows may not be as monolithic as it seems. Internally, there are many layers and boundaries. But, it is delivered as a monolithic product, and because of that, organizational dependencies have been allowed to remain. High level application changes, UI changes, and kernel changes all end up on one huge Gannt chart. Despite attempts to avoid them, true development dependences exist between layers that simply should be separate.
What if, instead of a 50 million line Vista, we had the following:
- A Windows TNG Kernel OS which has a separate release cycle, is developer-configurable, and was used in both desktop, embedded, and special-purpose applications by Microsoft and third party developers. It would have drastically simplified security, a lightweight process model, and could be built in custom configurations by developers (similar to the way Windows CE is delivered). Small configurations may have only a 400K footprint. Large ones would have no larger than a 1.5M memory footprint. (1-3 million lines of code)
- A Windows TNG UI (essentially "Avalon-in-a-box"). Again, a separate release cycle, and custom-configurable. This is a developer's product. Because it is separate from the kernel, competing UI models can be developed and delivered separately. Much of the "compatibility" with older Windows applications lies here. (3-5 million lines of code).
- A set of UI applications bundled with installers which is what users see as "Windows". It installs and runs very much like Windows does now. It is completely separate from the other two components, has its own schedule and can even be purchased in its unbundled form. This is especially appealing for corporates who may want the important security or functional benefits of a new kernel but do not want to retrain users until later. (5-10 million lines of code).
- The OS is unbundled, configurable, and separately shipped across all versions.
- The Microsoft UI is optional, but shipped with the consumer product as standard.
- On embedded devices, there is a more compelling case for vendor-specific UIs.
- Specialized embedded devices, such as wearable devices, benefit from a very lean kernel with no GUI overhead.
Most importantly, all variants use the same kernel.
What happened to the other 32-41 million lines of code? Well, they may exist somewhere, for example:
- A product that might be called "Windows Legacy". It's a virtual machine that runs a version of "Vista Minus" (or even "XP Minus"). Maybe it comes free with Windows, and is even installed by default for a while. It will run 100% of all old Windows applications. Thought it provides compatibility, it also sends a strong signal to customers that there is a dividing line beyond which compatibility may not be guaranteed in the future, and it serves to define where that line is at.
- Specialized Server applications such as IIS, remote management software, etc.
Most of the savings comes from the next goal...
Goal 2: Reduced Complexity
Windows has become far too complex, and needlessly so. For example, the past 15 years have seen 4 different phases of graphic support:
- The pre-NT GDI model present in products such as Windows 95 and Windows 98.
- The user-mode GDI model present until Windows NT 3.51.
- The kernel-mode GDI and GDI+ model of NT 4.0 and XP.
- The new Avalon framework for Vista.
While all of these are improvements, the compatibility requirement means that the legacy of these four models will exist for a long time. This makes the product tremendously complex, holds back progress, and adds to the number of details applications programmers must learn.
Similarly, the Windows security model is orders of magnitude more complicated than Unix-based systems. While Unix-based systems clearly are overly simple and not a good model either, the Windows model provides a detailed object-level security model which is far too expensive and yields few real benefits. Rather than being used by developers, most security settings on most objects created in the Win32 API are left unchanged. On a more pragmatic level, the weak passwords of most Windows users, coupled with the tendency of most non-corporate users to run as Administrator have rendered any security scheme irrelevant.
Vista is about to add yet futher layers of complexity on top of the already existing layers. Localization, security, object management, and graphics are all about to be reinvented in Vista. And yet, the old way will still remain.
Reducing the complexity of features such as these is essential, especially if upward compatibility is important. The worst thing you can do for future systems is saddle them with complicated features that are superceded and replaced constantly as versions are released.
Reduced complexity can also have performance benefits. For example, the Windows process model has always been criticized as "heavy". Creating a process takes at least 10 times longer than creating a thread and frustrations over a complex process model can hamper efficient use of resources. A new process model where threads and processes have the same weight, and processes can be created cheaply and easily would create greater performance opportunities in server-based applications. This becomes especially important in real-time and embedded applications where lightweight processes are almost essential to development.
What about Compatibility?
Compatibility is one of the most difficult constraints in moving forward. While Microsoft wants to say with 100% certainty that "Your application will run", the continual requirement of compatibility hampers progress on newer, superior technologies. In addition, compatibility makes Windows itself less flexble. If compatibility issues extend deep into the kernel and user APIs, then trying to deploy the Windows kernel in tiny embedded products will be almost impossible.
Microsoft will have to draw a dividing line in the sand, as Apple has done. Applications on one side of the line will run with few if any changes, and this should represent about 90% of the applications developed in the past 10 to 15 years. This means that some kind of "compatibility library" will need to be developed.
Isn't Microsoft Already Doing This? Why start over?
Well, yes, and no. While many features in Vista are targeted at these problems (such as the "Server Core" version of Vista), there is no true unbundling, and Vista will remain a monolithic product. It will take years, or even decades, to gradually pull the pieces of Vista apart, and while those attempts are made, developers will continue to add more. Attempts by Microsoft to truly create a layered operating system out of XP with no layering violations have been difficult.
Make or buy?
In many ways, what I'm suggesting is obvious. Microsoft knows they'll have to replace Windows. That's why projects like Singularity exist. And, the goals I've spelled out above (including unbundling) may already be on the drawing board.
In practice, the sheer size of Windows, and the compatibility juggernaut, will make everything take longer. If Microsoft wants to replace Windows by 2015, it will take until 2025. If they want it to be "fully compatible", then even Singularity will be hampered by the very same issues outlined in my previous post and this one.
Can Microsoft, and their customers, wait until 2025 until repeated evolutionary steps solve all the problems I've mentioned? Will Microsoft continue to have such dominance that people will wait? Unless Microsoft acts more quickly, it is inevitable that some competitor, probably Apple, will finally be able to attract large numbers of Windows users with an offering that, to users, appears similar enough to Windows from a purely functional perspective.
Maybe it's time for Microsoft to do what they've done so often before: Acquire technology which solves the problem.
Here comes the L-word
Everything I've said until now has tried to make a case that:
- The liability of the Windows codebase, including Vista, will slow Microsoft's progress to the point where vulnerability to competitors becomes threatening to Microsoft within the next 5-10 years.
- Microsoft is solving these problems, but not fast enough for their users or shareholders.
- It's worth considering if there are potential technologies which can be aquired to solve the problem.
Linux has the potential to solve Microsoft's problems, but it's important to look at the potential of Linux rather than the current reality. Consider that Microsoft Visual Basic was originally purchased by Gates as Tripod by Alan Cooper, and SQL Server was written by Sybase originally until Microsoft negotiated exclusive rights to all Windows code. What those products are today only a slight resemblance to what they were on the day they were purchased.
So, rather than look at Linux the way it is today, imagine what it would be like if Microsoft were to adopt the Linux platform, participate in and fund development, and drive the direction of Linux forward to meet Microsoft's own needs.
Technically, Linux has the following strengths:
- It is a successful platform in use today, which is benchmarked and compared side-by-side with Windows. In server-based applications, it often comes out ahead of Windows in some performance and security benchmarks. Rather than being an "idea", it is an actual contender and is being used side-by-side with Windows in many corporate production environments.
- Its security model is remarkably simple and even could be called antiquated. Yet, for some reason, it has held up very, very well and is considered at least as secure as Windows. Because it is so simple, it will be easy to upgrade and replace. But, because it is working adequately now, replacement can be done carefully and cautiously while focusing energy on more important issues such as desktop innovation.
- It has an efficient lightweight process model that is a superset of the one provided by Windows (that is, Windows process model can be built on top of the Linux process model).
- It has been competing vigorously with Windows, and there is already a large device driver base. In fact, vendors of hardware consider Linux to be second only to Windows in their priority for releasing device drives, and many vendors already do.
- It has an entrenched development model which is popular in universities and many businesses. thus, it is not necessary to spend significant time or effort on development products, especially for low-level drivers and server applications.
- It has a configurable kernel which can be used in everything from tiny embedded devices up to very large multiprocessor systems. The kernel is small, modular, and extremely robust.
- It is much newer than Windows, and has very little legacy code by comparison.
- An enormous amount of effort has already been done to create Win32 compatibility layers. WINE, Crossoffice, and Xen are three specific technologies designed to run MS applications under Linux. Rather than criticize these as inadequate and lame (they are), consider what would happen if Microsoft were to take over development of one of these. Progress would be rapid and the problem of compatibility would be cleanly isolated.
- It has a shared library model which allows multiple concurrent and incompatible versions of software to co-exist simultaneously without the need for extensive additional technology investment or developer education.
Linux has several weaknesses to consider:
- The X-Windows platform is interesting, but outdated. While a client-server windowing system has clear advantages, its API is more arcane and complex than Win32. Microsoft (and the market) would be served well if Microsoft were to build (or buy) a new, more modern and capable graphics application framework for Linux.
- The Open Source model requires a substantial investment in legal work and planning. While some of the technologies would clearly remain Open Source, Microsoft would want to engineer as many components to be proprietary as possible. This may dictate packaging and delivery "mechanics" in some ways. Since many companies are already combining Open Source with proprietary products in their deliveries (such as Redhat), I am confident Microsoft can negotate this minefield in an aggressive and innovative way that would impress us all.
But, the biggest benefits of adopting Linux aren't technical at all...
Linux: The Business Reasons
While the technical issues can be argued to death, the business reasons for adopting Linux give Microsoft significant advantage.
Keep in mind that Microsoft has always been excellent at the adoption and assimilation of technologies that are already in the marketplace. I remember when the Web was something Microsoft said they weren't interested in! (Yes, it's true). Despite the Java legal debacle (which Microsoft could have avoided, I believe), their adoption of Java was highly successful and if managed properly could have avoided having to "reinvent" replacements such as C#.
So, here are what I believe are the most compelling reasons why Linux is a good "buy" choice:
- Microsoft completely eliminates open source as a competitor. By embracing Linux, almost all open source efforts suddenly lose most of their "shared mission" to compete with Microsoft.
- Microsoft extends dominance of the Office applications onto every desktop. Efforts like Open Office, Sun Office, etc. become truly jokeware.
- Competitors like IBM, RedHat and Sun start to shake in their boots as they realise that multi-million dollar investments they've been making in Linux have now directly benefited their most feared competitor.
- Microsoft takes firmer control over how the GPL is applied to products. Keep in mind that many products you buy contain both open source as well as proprietary components. Windows TNG would be no exception. But, because it was bread-and-butter to MS, they would apply their significant legal capability more productively.
- The Open Source "religion" would become diffused. Most open sourcers would be horrified to think that MS is "taking over". As more and more MS successes occurred in the Open Source arena, the "we love Linux and software should be free" crowd become more and more marginal.
The Bottom Line
Sure, it's a minefield. But, Microsoft is on a slow-burn right now to creating less and less competitive products while others are creating more innovative products with shorter delivery times. By unbundling, creating leaner development strategies, adopting some proven technologies, and dominating the open source space, Microsoft can reinvent the entire industry.
That leaves Microsoft more time to outdo the very competitor who is making the greatest advances toward Microsoft's market: Apple.
The next segment of this post will explore how Microsoft might refocus and use their time to create a truly next generation desktop by creating a proprietary application layer on top of Linux and OS/X.