Why Linux May Never Be a True Desktop OS

This week I’m speaking at the Desktop Linux Summit in San Diego on why the PC OEMs don’t, and probably never will, fully support Linux on the desktop. This is somewhat of a deja vu for me since a decade ago my team made a similar presentation at IBM on OS/2 and why it wouldn’t beat Windows in its time. For the purpose of this piece I’m using the word “Linux” as an open source OS catch-all so I don’t have to call out every distribution or variant.

Now, as it was then, the core of the argument wasn’t the technology, the strength of Microsoft, or anything, really, outside IBM’s control. The reason was that IBM would simply not do what was necessary to take the market away from Microsoft.

When I look at Microsoft these days I recall something former Hewlett-Packard Chairman Lew Platt once said: “My biggest problem is finding people who will tell me the truth.” I’ve never seen Microsoft as weak, from an execution standpoint, before, but currently the firm appears weak. Taking advantage of that weakness remains elusive for both Linux players and Apple, however. Let’s take a look at the reasons why it’s been difficult for them to move dramatically into the business or consumer desktop computer spaces.

Doing What Is Necessary

When it comes to competition, typically you have the folks who are willing to do what is necessary to win and the folks who aren’t. In my personal experience, I’ve seen again and again scenarios where a team would lay out a plan for what they deemed was needed in order to be successful in a market and then they’d watch as the executive in charge cherry picked the things he/she wanted to do. As a result, the project failed, often disastrously.

People like to put filters over their eyes and ignore important steps they don’t want to perform on the road to achieving a goal. All of us have seen examples of when this derails the plan. With OS/2, the last major challenger to Windows, IBM had a list of things it needed to do in order for the product to be successful. This list laid out staffing levels, marketing expense, structural changes to the company and partnerships, and its foundation was one of the most extensive pieces of research that company ever did.

However as these recommendations went up the chain of command they were changed or in some cases, ignored. The end result? OS/2 failed.

The requirements for OS/2’s success included broad third party hardware and software support, developer support, and Windows 32-bit support. In order to get into the hardware, however, OS/2 couldn’t come from IBM. This was probably the single biggest deal breaker. Compaq, HP and others wouldn’t support an OS from a competitor and, much like it was with Sun and Java and Netscape and Navigator, the biggest weakness that OS/2 had was IBM not doing what the market required of it to be successful.

After reading up on a subject that speaks to a likely OEM desktop requirement — strong vendor supported drivers — I deeply doubt whether the folks working on Linux will make the difficult tradeoffs needed to assure success. Think of the iPod, for example. In order to achieve success Apple had to support Windows, as distasteful as that must have been for the firm.

The OEM Cost of Linux and Windows

Windows is free to the OEMs. In fact, not only is it free, but Microsoft, in effect, pays them to take it. Regardless of the cost, Windows is a logical choice, and a straight pass. Dell pays about $80 for it and typically charges about $80 for it. There is rarely much of a mark up. If Microsoft were to lower its price that lowered price would be reflected in virtually all desktop hardware immediately.

Microsoft provides a number of services which include development support, service support, marketing support, technicians, classes, databases and support materials, and it picks up a lot of the service load as well. In addition, it provides marketing co-op dollars, incentives for early adoption of new products, and ensures a somewhat level playing field (which could be good or bad) for the vendors.

Linux, on the other hand, is often viewed as “free.” Even when it comes with a fee, that fee is also passed through without any mark up. However, the hardware vendor must now provide all of the things that Microsoft has been providing, too. Particularly painful are services, as customers expect a similar experience with open source that they have with Windows and will often call the vendor to help troubleshoot. The vendor takes all of the related costs; there are typically no dependable (on-site) resources, unless the vendor supplies some of its own.

With Linux, the customer often expects to get the product for free and wants the retail price of Windows deducted from his/her purchase price. There are no funds passed back to the vendor and, because Linux is different, customers tend to place more service calls — at $85 a call. As a result, the vendor generally ends up losing money on average with Linux.

This is the primary reason vendors don’t want to do desktop Linux, unless there is an extremely lucrative services contract tied to it. Given services revenue often flows to the services division, the desktop hardware divisions still do their best to throw their bodies in front of these efforts.

Too Many Versions

An old rule of the desktop is that support costs go up by the square of the number of different products you have to support. This may actually understate the problem now that much of this support is done in remote sites like India.

For each version of a product the support team has to be trained, support scripts have to be created and updated; patches have to be collected, tested and deployed; and line changes have to be tested against every operating system version shipped. This is part of the reason that IBM hardware was not competitive in the early nineties; the extra cost of supporting OS/2 was theirs and theirs alone while IBM software got the largest portion of OS/2 revenue (that division was profitable, though OS/2 wasn’t).

There is broad diversity among Linux buyers, and Red Hat, which is the closest to a standard Linux OS, isn’t yet dominant enough to claim victory. The OEMs might be able to handle one version of Linux but until there is only one they have to handle they will probably avoid doing it.

Excessive Advocacy

A few years ago I recall getting a note from an Apple guy who said something to the effect of, “I’m sure glad all of the crazy Apple supporters moved over to Linux.” While I seriously doubt all of the crazy Apple zealots moved to Linux, I can say that they were instrumental in getting Apple out of large enterprises.

Zealots, regardless of cause, have a tendency to treat very poorly those who don’t share their beliefs, and, historically, they have been much more likely to misuse e-mail or take a dispute to a public forum. They tend to be human resource nightmares and anyone wanting to create a powerful argument against either platform has only to take the decision maker to any of a number of fan Web sites to create the impression of significant risk.

This can be particularly problematic now with employee blogs starting to spread and more internal disagreements being made public. Management typically does not want to create an environment where company dirty laundry is aired or where employees organize to fight executive orders. Yet zealots appear to do this as a matter of course: their beliefs come first, even though this is technology, not religion.

With all this in mind, IT executives want to be able to decide to abandon a Linux or Apple trial with faith that the employees would follow those orders and not revolt. They also want to feel in control with regard to how the technology is used and not cede that to an external or internal employee advocacy group without conscious thought.

In conclusion, there are a few key requirements for alternative platforms to be successful, success being measured by significant market share, in the general market.

A Reliable Road Map to Linux Desktop Success

The list of clear requirements includes:

• Clear OEM resources to match or exceed those currently provided by Microsoft. This would include R&D support and co-marketing dollars.
• Embrace existing desktop requirements (roadmaps, application support, proprietary drivers, consistent patch releases schedules and documentation, clear escalation lines for support).
• Seek out reasonable advocates who will take direction from their executives and not from other advocates.
• Make it profitable for the OEM.

Recall that for Microsoft it took about 5 years, US$10 billion and Y2K to move the majority of the installed PC base to another kernel — and they had to kill the old one off to do it.

In the end, I look at this short list and still doubt it is reasonably achievable. I also wonder if Linux were actually modified to be able to comply with this list whether it would even be Linux anymore. That deserves a little thought. Generally speaking, you have to give something up in order to get something else. However there are times when what you sacrifice is actually more important than what you gain.