The Free Linux Open-source Threat
The essence of your argument is that open-source opens the door for inexperienced vendors to offer quality software cheap. You contend that such vendors make for inadequate support, especially since their business models stink and soon most will be out of business.
While I agree with your position, you should know there is far more substance to the open-source movement than you seem willing to suggest, but far less a threat to WIND than even your pessimistic view of the strength of the movement implies.
By definition, free open-source means a developer can hope to gather up for zero cost the source code for all software needed for a project. This includes the OS, compilers, debuggers, protocols, and middleware. The purest will download all, or most, of the software from numerous sites, obviating the dependency on any vendor whatsoever.
If you have thoughts of creating a roll-your-own software OS and development suite, open-source Linux software is a splendid place to start. Most of the software you will need can be downloaded free with source, and the common Linux API makes for a relatively simple build, at least for a supported hardware platform like the Pentium. And you are vendor free!
If you worked for me, you would also be job-free. The notion that that there is virtue in being totally sufficient in all aspects of product development is absolute nonsense. The modern world benefits from faster and faster technology development, while being overwhelmed by escalating complications. The solution has been obvious for years, and it consists of outsourcing everything but core competencies. Nowadays more and more companies are outsourcing everything but their concept, literally becoming virtual companies. Semiconductor Intellectual Product companies, like Rambus, ARMS and Mips, do not make chips, only designs. Many chip makers don’t own fabs, choosing to out-source all manufacturing. Contract manufacturing, particularly of electronic products, is booming, and shows no signs of slowing down. Cieva, the electronic photoframe company is about as virtual as you can get; it farmed out the design, development, and manufacturing of its product, and sells and maintains the product over the web, the development of which it probably farmed out.
If anyone had the resources to go completely in-house and sufficient in software development, it would be Intel or Sony. Both companies are moving in the opposite direction.
The problem with open-source is the open source. This empowers programmers to tweak the code to improve or add a feature, or to port the software to an alternate hardware platform. With millions of lines of code, you have now found and altered 5000 lines. What are the chances that you got it right? What are the odds that your intrepid programmers are properly up to the boring task of documentation and configuration management? What are the chances that future changes by different software engineers will be compatible? What will you do when a sweeping hardware upgrade occurs, like SOC’s or Network Processors, and your ability to even understand the hardware much less tweak millions of lines of code is zero to none?
I am not surprised that technical trade magazines and the press rave on about the empowerment of open-source, but don’t let these pundits fool you. Successful managers and experienced engineers will not suffer these fools. The open-source movement will be swept to the sidelines of mainstream product development by the vastly more powerful movement toward ever-increasing outsourcing.
As the world enters the post-PC era with smart, connected devices everywhere, there is a growing need to somehow come up with, not just billions of device units, but millions of device designs. If everyone on earth became an embedded software engineer working with open-source, we would lack the manpower resources to meet the developing need for smart devices. Like it or not (and the prototypical embedded engineer will not like it), but the future of product development will depend more and more on ultra high-level development tools resting confidentially on top of a mountain of integrated code. The embedded engineer of the future will have no more knowledge of the details of underlying software than does today’s enterprise application programmer know of the details of TCP/IP, database internals, or what makes a compiler compile, not to mention the merits of alternative kernel architecture.
As an example of the future, WIND’s TDI claims Honda developed ECN’s for a future car directly from MatrixX without any programmer intervention. E-Sim makes the same claim with its finite state machine toolset called RapidPlus. These are exactly the kinds of tools that will be needed to meet demands on future designs, but all such tools depend on everything below them working perfectly. No glitches can be suffered. In the face of any integration problems whatsoever, high-level programming tools are worse than useless, because they introduce complicated machine-generated abstractions which compound the difficulty of low-level debugging. This means, the software foundation incorporating all needed connectivity protocols and application-support feature-sets must be available and trustworthy for any desired hardware platform. None of this is practical with open source.
This is why WindRiver gets the big bucks and always will.
Allen
PS -- I don't mean to imply WIND has filled every gap and obviated any need for source code. Indeed, WIND is compelled to provide source code when it sells protocol software because the code obviously is intended to be integrated into a software bundle. The primary winds of change, however are pulling the company toward integrated software solutions, diametrically opposed to open-source. The irony is that WIND is a major beneficiary and contributor to the open-source movement. |
