|
|
Guest editorial: Kevin Dankwardt responds to EE Times editorial
2004-01-19 Article Rating:  / 1
Foreword: The following open letter to the EE Times responds to this editorial by Green Hills Software CEO Dan O'Dowd, in which O'Dowd denies the existence of a market for embedded Linux development tools.
The letter was written by Kevin Dankwardt, a respected embedded Linux instructor and LinuxDevices.com contributing editor. As background, you may also wish to check the talkbacks posted here. January 19, 2004 To: Brian Fuller, Editor-in-Chief EE Times Dear Mr. Fuller: Dan O'Dowd's article entitled "The myth of the embedded Linux tools market" (EE Times, January 12, 2004) is deeply flawed. As a professional educator of developers in the embedded Linux market, I have concerns about Mr. O'Dowd's allegations and conclusions. Ample contrary evidence can be cited to challenge O'Dowd's statements in nearly every paragraph. First, there is no myth as alleged in the title. Just google "development tools for embedded linux" to prove it. We have grown used to negative opinions about embedded Linux coming from those most threatened by it. Most contra-opinions issue from Redmond, which makes O'Dowd's view an attention-grabber. After all, O'Dowd and his company have been an embedded fixture. The article focuses on tools for embedded Linux developers but also refers to the general use of embedded Linux. With regard to deployment of embedded Linux we can simply look at the market and see that the technology has become the majority platform for embedded systems. The main reasons include easy and free access (obtainability), freedom from royalties, source code availability and vast global community support. The "roll your own" crowd -- typically about 50% of the market -- now starts new projects with Linux more than any other OS. Linux delivers an acceptable size (footprint) and performance profile for many, many embedded devices and applications. There are several different commercial hard real-time solutions for Linux as well. Just because some vendors don't offer hard real-time (or confound the issue through their use of the term real-time) does not mean that Linux is not capable of real-time performance. Much of the argument revolves around characteristics of Linux. For example, the fact that Linux may need a RAM footprint of 5 MB or more is somehow deemed unacceptable in an era of cheap bootable flash disks many times that size. What percentage of powerful embedded devices that require a sophisticated operating system can afford such a footprint? I'd say nearly all. Just because Linux would not be a viable option in twenty year old devices does not mean that it is not viable today. The times and devices have changed. It's clear to developers: Linux is extremely suited in many dimensions for most of today's sophisticated devices. A major reason that device developers are switching to Linux is its power and wealth of capabilities. Mr. O'Dowd claims that other "equivalently configured" embedded operating systems may be 10 to 20 times smaller. Equivalently configured is left to the reader to define. Are these other operating systems, say those that have a flat memory model, even capable of being anywhere near equivalent to Linux? O'Dowd also points to the diversity of embedded Linux distributions as reason enough for tools vendors to be incapable of making a profit. This is the so-called myth, and it's a lesson in contorted logic. Tools must run on a host first and only then be concerned with the target. On the host side, if a vendor supports Red Hat Linux they have covered the majority of the embedded Linux developers. In my personal experience, this is true for developers worldwide. As for targets, it may indeed be inadvisable for vendors to attempt to sell tools that have significant dependencies on target peculiarities. However, a virtue of Linux is that software can be amazingly portable, at the source code level, across architectures. Thus, tools that work at the source code level, or on the host, find wide applicability. And that's no myth. In addition, the Embedded Linux Consortium's platform specification (ELCPS) is serving as a clear standard for developers regarding the functionality to expect on targets. The ELCPS provides three resource levels ranging from low to high. And, the ELCPS is destined to mature with the market. The majority of O'Dowd's article revolves around developer tools and the business model of tool purveyors. The article implies that an important feature for embedded developers is that the tools should lead to smaller or more efficient implementations. This is clearly a narrow and backward-facing definition of the role of developer tools. While this role may have had support in the proprietary, closed past with seemingly insurmountable hardware limits, times have most definitely changed. To conclude as O'Dowd does that there is no market for tools because the old business model no longer functions in an open source world is just backward facing logic. Just as Linux has proven to be a disruptive technology in operating systems, open source has proven to be a disruptive technology for software in many areas including development tools and methodologies. Many vendors have created successful new business models. Revenue success has been achieved, for example, even when the product (the operating system) is given away. Let's continue to look forward, seeking more and better tools for Linux. Yes, the need is there for improved tools. It's always been there, in every electronic engineering market. Many of the new tools are likely to be open source (if recent history provides guidance), and the financial success of their backers will revolve around support, funded enhancements, training and facing the right direction. Kevin Dankwardt President, K Computing Education Chair, Embedded Linux Consortium k@kcomputing.com Do you have questions or comments about this story? Talkback here! Related Stories:
|
news feed
