Developing Linux for the desktop market


Yankee Group Research Fellow Laura DiDio, who specializes in software platforms and app infrastructure, acknowledges the strides made by desktop Linux distributors, such as Red Hat with its latest release, Fedora Core 5.The desktop decision, however, requires giving orgs where Windows is entrenched a reason to move. DiDio discusses her outlook for desktop Linux with Senior Editor Kathleen Richards.

ADT: How well entrenched is Linux in the server market?

DiDio: Linux at this point globally has about 20 to 25 percent market share in the server market. I think in Europe it is at the higher end, 25 percent. It has certainly been on a steep ramp but coming from a very small base the trajectory is going to be almost vertical. Certainly Germany, UK, France in Western Europe really like it. It is getting a lot of mainstream acceptance and a lot of buzz around it in Asia Pac as well and here in the US. Those are the top markets right now.

ADT: What about on the desktop?

DiDio: Well the desktop has had much lower adoption both from unit shipments and market share percent-wise. At this point Linux on the desktop is maybe 2 to 4 or 5 percent worldwide.

ADT: Is that because Linux is still only appropriate for narrow applications?

DiDio: No, not at all. Linux can be adapted to many applications. The appropriateness of it is going to depend upon what an individual organization wants to do and the capabilities that they have internally to support it. You also have to make the distinction between commercial Linux—Red Hat and Novell SuSE desktop Linux—from open source desktop, things like Linspire, etc. The difference is that you will pay a licensing fee for commercial desktop Linux, the same way you will for commercial software. Of course, commercial Linux, you are only paying for one license. You are not paying for client access to the server like with Microsoft. In that sense, it is a lot cheaper.

But at this point and time you don’t have the full ecosystem built around desktop Linux. I’m talking here about a very broad and deep range of applications that are supported, management tools and utilities, performance enhancement utilities, lots of documentation, knowledgebase articles, like you have with Microsoft with TechNet. All of those things. Plus there are still not as many Linux consultants or skilled IT managers—and I emphasize skilled here—who know all the ins and outs. People will argue, ‘hey Linux can be easy to use etc., if you have the right skill set,’ but there are going to be integration and interoperability concerns that customers have to be concerned with. The other thing is companies need a reason to migrate from something that they have now to Linux. What are you going to get out of it? There are a lot of things that you can do easily with Windows today that would be harder to do for the average corporate entity such as remote access.

ADT: What about the argument if you are using Windows 95, and paying for support, migrating to Linux instead of the next version of Windows might make sense whereas if you are using later releases of Windows, it is a different business model?

DiDio: Very few people still have Windows 95. I think if you are using Red Hat Linux or Novell Linux, one of the big commercial server solutions, that it makes it easier to take the next step and go to the desktop, if you are getting a lot of service and support from those vendors.

ADT: The major Linux distributors are supporting the Linux Standard Base 3.1 specification, which was announced in April. It is the first LSB spec to include desktop app support, purportedly making it easier for developers to develop for a core Linux platform because they will be supporting common software libraries, for example. Is this expected to have an impact?

DiDio: That will certainly help the open source and Linux development community. But that’s not the issue. The real issue is how do you dislodge the entrenched desktop operating system, which for 90 percent of users is Windows. You can’t just say, ‘okay yes Linux might be cheaper.’ The initial capital expenditure costs on licensing for client access licenses might be cheaper than Microsoft, but Microsoft bundles in for the price of that license a lot of freebies and a lot of training, service and support, home usage rights, discounts for employees of up to 40 percent on other applications. They also give you a very comprehensive product warrantee.


What I found from my recent reliability survey, much to my surprise—I surveyed 400 companies globally—50 percent of the people who are using Red Hat and Novell SuSE, are customizing the server releases, which would indicate that they would probably do some customization of the server and desktop apps as well. Customizing Linux is a two edged sword. The downside is, it’s an application and if you customize it, guess what, none of your vendors will give you any product warrantees, if something goes wrong. So companies that have a very low tolerance to risk might find themselves going out and buying outside insurance.

Companies that elect to use an open source desktop are going to have to assume the responsibility for managing and maintaining that, and going out and getting their own update for things etc., looking for some documentation. If you have the type of environment where you have a lot of engineers or gurus or very skilled software developers who know the Linux and open source model well, you’re golden. It can be very low TCO and a near immediate ROI. It is going to depend on the organization’s particular situations.

ADT: Are more Windows ISVs starting to develop for the Linux environment?

DiDio: Yeah, but I think the really interesting question is, are they backing off the Windows platform? And the answer to that is they are not. You can track when Novell Netware started to lose market share, it was when the application developers found that they couldn’t make a lot of money. Anybody that is developing third party applications for Linux is going to be charging you the same amount of money for those Linux packages that they are charging for Windows, Macintosh and Unix. Those third party apps, tools and utilities are not going to be free.

ADT: What is your opinion on mixed environments? Are they cost-effective?

DiDio: The research that I have done over the past four to five years indicates that over 80 percent of organizations have heterogeneous environments and most on average have three different servers in their environment. Now that might be an Apache Web server or specialized server, they might have Unix in the data server and Windows for the line of business. In some cases what we are finding is Linux is nibbling at the edge, getting installed as a specialized app server, or maybe in a certain department. To date, most of the defections to Linux have come at the expense of Sun Solaris on the server side.

Do I see Linux getting a lot of traction at the desktop at this point? No. Certainly some, but we are not talking about the migration of the wildebeests here. What we are finding is the fat client Linux desktop is still best suited for the technical engineering development environments, where the corporate organization only has to depend on a native Linux app, or a subset of Windows apps that might be running.

The point is you need a reason to move, more than a reason to stay. And Microsoft continues to make a lot of improvements to the Windows environment. We also need to see broader hardware and OEM support for Linux. Out the box, many of these Linux distributions require customers to do additional configuration in order to work with different types of vendor hardware, particularly things like the video and audio drivers. Networking and WiFi configuration are still not plug and play.

I think where Linux has the best chance is in greenfield networking situations, where there is not an entrenched operating system, which, is why you now see Microsoft cutting prices or coming out with light versions of Office or XP. So, for instance, an organization in China that doesn’t already have Windows on the desktop.

It is going to be a tougher sell, dislodging what is already there. Certainly, we’ve seen a lot of municipalities like the city of Munich say ‘okay, we are going to switch out our 1400 desktops and we are going all Linux. And other groups say, ‘not so fast, let’s do a business case on it.’

Any piece of software or any combination of software may be suitable for a particular business as long as the business can make the business case. And that means doing a thorough TCO analysis. In many cases, we see companies who can’t move from one version of Windows to another, or one version of Office to another, because, guess what, they have a specific vertical application that is not yet supported by the new environment, and that might be true of Linux. So usually it is the killer app which really dictates whether or not an organization moves forward, sticks with what they have, or rejects something new.