Foreword: This interview with Denx founder and CEO Wolfgang Denk discusses the popular and free Denx embedded Linux distributions, Denx's business model, software patents, and other matters related to embedded Linux markets and technology. It is the sixth in our “CEO Interview” series. Enjoy! . . .

---

Q1: Can you say something about the Denx distribution, which seems to be fairly popular?

A1: The project for our Embedded Linux Development Kit (ELDK) was started to keep Denx and our customers independend from third-party, usually more or less proprietary solutions. As nearly all our work (including the ELDK) is based on Free Software, we consider it only consequent to make the toolchain available for free — completely, i. e. with all patches and tools and everything.

When other vendors increased prices for embedded distributions, Denx considered it a business chance to go exactly the opposite way: make all the software and tools available for free. So far this has worked very well.

The ELDK is available for PowerPC, ARM, and MIPS targets, and thus covers all major embedded platforms.

Q2: What's the basis of your business model?

A2: In short: we sell our time and experience.

Looking at the big Linux vendors like RedHat, SuSE, MandrakeSoft etc., it seems really difficult to make profit by selling Free Software as a product; if these companies make profit at all they do so by selling services. So this is what we do — and have been doing right from the beginning.

This works pretty well for a small company like Denx: we are a total of five engineers in Germany, backed by additional man-power outsourced to some freelancers and a company with excellent engineers in Eastern Europe. Our turnover reached more than $ 1.1 million in 2003, growing 77 percent in 2001, 32 percent in 2002, and 66 percent in 2003. We have customers not only in Europe, but all over the world — from the U.S.A. to Japan, from the United Kindom to Australia; we are proud that our list of references includes telecommunications companies like Alcatel, Ascom, Lucent, and Siemens; processor manufacturers like Infineon and Motorola; board manufacturers like Force Computers, Microsys and TQ; as well as other big names like DaimlerChrysler and Liebherr.

Q3: How would you characterize the current state of embedded Linux, as a market and as a technology?

A3: The employment of Linux as an operating system for embedded devices has become standard technology. We can see this clearly when talking to our customers: four years ago engineers asked if Linux would work for their project; two years ago they asked us for arguments they could use to convince their managers; today, they tell us that their managers required that they switch to Linux.

Q4: What challenges do embedded Linux OS, tools, and services vendors face?

A4: A big challange is fragmentation. For example, just look around how many different Linux kernel source trees there are for PowerPC systems; there is:

linux-2.5 — kernel.org official 2.5 tree

linux-2.4 — kernel.org official 2.4 tree

linuxppc_2_4_devel — old PPC development tree

linuxppc-2.4 — PPC patches not yet sent to linux-2.4

linuxppc-2.5 — PPC patches not yet sent to linux-2.5

linux-2.5-ocp — PPC patches for 4xx OCP devices

benh — Benjamin Herrenschmidt's PMAC stable tree

linuxppc_2_4_mpc5200 — old MV tree including support for MPC5200

linux-2.4-mpc5xxx — newer tree including support for MPC5200

linux-2.4 @ ameslab — 2.4 ppc64 tree

linux-2.5 @ ameslab — 2.5 ppc64 tree

Then there are more or (usually) less open trees from several vendors (MontaVista, TimeSys, FSMlabs, Sysgo, Denx, …) or from some consortiums (like the CELinux source tree). Then there is a great number of kernel patches, and even of bigger patch collections like the Linux-Tiny Tree.

How would you answer the simple question: Which source tree should a developer use as the base for his next project?

It is obvious that many different groups are working on different (and quite often on the same!!) issues. So far, existing organizations like the ELC or CELF have not changed this at all. It seems also obvious to me that many companies are not interested in sharing their results with the other developers of the Free Software community.

A lot depends on Free Software remaining Free Software — which boils down to how much the GPL is worth if tested in court. I was really happy when I learned that the iptables/netfilter project was able to have a restraining order placed against a router manufacturer violating the GPL.

Q5: What market opportunities do you see for Linux in the embedded devices and systems market?

A5: Linux is winning ground in all segments of the embedded market. It has become a very strong competitor with all other RTOSs — proprietary and “open source” as well.

But Linux is especially the primary option for many projects where so far self-written simple schedulers or one of the zillion unknown RTOS was being used. This way Linux is enabling especially small companies and smaller projects in big companies to use high-end technology for their products. It's really fun to see how our customers discover the boundless possibilities of their new environment.

Q6: What challenges does embedded Linux face as a technology?

A6: Software patents are a big threat that challenges Free Software as a whole, not only Linux. In the USA, this already has been a problem for a long time — see for example the RTLinux patent issues. So far, we in Europe were much more lucky because we have much stricter rules for granting patents. However, the European Council of Ministers is covertly pushing for unlimited patentability of software right now. Denx takes part in an online demonstration and has closed our website in protest.

Q7: What embedded Linux technology developments do you find exciting?

A7: This is difficult to say. I'm especially fascinated to see the designs of our customers, usually long before they enter the market. For me this is actually more exciting than the technology we use ourself — embedded Linux is just a tool after all.

One interesting development is to watch how more and more processors come into life which are specially designed for embedded systems, and which have all the required resources to run a standard Linux — like the Infineon INCA-IP VoIP engine or the TI OMAP as used for example in mobile phones.

Q8: Can you share one or two of your company's most exciting successes?

A8: I think one of the most successful projects we did (and still do) is the Universal Boot-Loader (“Das U-Boot”). You can find it anywhere — in cigarette vending machines, in mobile phones, in Bluetooth and WLAN access points, in network switches, in mobile cranes, and even on board of the ISS (International Space Station). Started four years ago as a project for MPC8xx PowerPC processors only, it now supports all sorts of PowerPC, ARM, MIPS, x86, NIOS, ColdFire, and other processors, and more than 200 different board configurations.

And the project is still flourishing — nearly every day we receive contributions from other developers that help to improve and extend U-Boot with really powerful features. For example, we just added code for VLAN support — I don't think that many other boot loaders support this. It is really fun to maintain such a great project of the Free Software community.

One of our technically most interesting projects as a whole is the new generation of Liebherr Mobile Cranes. This is really a project where embedded Linux is running “under heavy load.” But then like in so many other cases — Linux is just a tiny building block in this machinery, and usually the other components impress me much more than Linux. As I mentioned before: Linux is standard technology today.

Q9: How do you see the Eclipse tools platform fitting into the future of embedded Linux development?

A9: Eclipse is certainly interesting for application development, but much less for the base software levels we are dealing with. Maybe it will be used in new projects, especially in Java based ones, but this is just a small section of the whole market. So far I am not aware of any significant free project in the embedded Linux community that includes direct support for an IDE. I don't think that Eclipse will change this quickly, especially not for existing and bigger projects – like for example the Linux kernel.

Q10: How do you see the 2.6 kernel affecting the embedded Linux market?

A10: There is more marketing babble than actual impact. Kernel 2.6 provides a couple of interesting features, but many of them are stable only on x86 systems. Kernel 2.6 is definitely a big advantage for MMU-less ARM systems: having uCLinux merged into the official source tree is a major step forward. On other architectures, like PowerPC or MIPS, 2.6 is not stable enough yet for use in real projects. And it's bigger, and sometimes not as fast as 2.4 — I think it will take at least six months before we can consider using 2.6 in a project for a customer. And we will support 2.4 for a long, long time after.

Q11: Any comments on the SCO mess?

A11: I really don't care much. I don't think they have any real proof for violations of their intellectual property — or proof that they actually own it in the first place. They are just spreading FUD. Their attack on Free Software is irresponsible — and desperate.

Q12: What do you think about Wind River's recent public endorsement and announcement of support for embedded Linux?

A12: Wind River can certainly become an extremely important tool and service provider. But whether they manage to tap all of their potential in the context of Free Software remains to be seen.

Wind River used to express a very strong anti-Linux position for a long, long time — maybe too long. We will have to see if their customers buy that their change of mind is heartfelt and honest.

Q13: What's your vision of the future for embedded Linux? That is, how big will the embedded Linux market get, and when will it start to reach that peak?

A13: At the moment I see virtually no limit for the use of Linux (and let's not forget FreeBSD, NetBSD etc.) in embedded devices.

Until now there has been a certain class of embedded devices which were simply “too small” to run a Unix based OS; but with the ever-growing processing power, more and more of them can — and will — run Linux.

Also, Linux is winning more and more projects where proprietary RTOS used to be employed traditionally — not all of the 90-to-180-degree turns of RTOS vendors like LynuxWorks or Wind River have been completely voluntary, I think.

---

About the interviewee: Wolfgang Denk graduated with honors in civil engineering from Ilmenau Institute of Technology. He has worked as a systems programmer and software engineer at Friedrich Schiller University, PCS Computer Systeme, Siemens AG, ICN (Information and Communication Networks, Transport Systems, Optical Networks). As an employee of Siemens, he was responsible for the OS for Optical Network Elements. He has done porting, driver, and release work with Unix since 1983. He founded Denx Software Engineering in 1999, and Denx Computer Systems in 2000. He says he has never worked under any sort of MS-DOS or Windows OS, “and never will, I guess.” Both Denx companies are 100 percent Microsoft-free.

---

 
This article was originally published on LinuxDevices.com and has been donated to the open source community by QuinStreet Inc. Please visit LinuxToday.com for up-to-date news and articles about Linux and open source.