Congratulations to all of you for contributing deep thinking about these issues, and that includes you, Ronald Paul. No flames to thinkers. This thread is responding to unfolding events in an informative, useful way. Let me try to summarize where we stand on a number of important items.
Jason is right, CTXS is not a useful analogy. Scratch all allusions to CTXS. (Jason's arguments hold. Ronald, if you disagree, we can discuss it, but I think the analogy should be dismissed.)
A key issue has been raised, by Ronald, about the lack of barriers in flipping from one RTOS to another. My sense is that organizations that commit to a particular RTOS and associated toolset will find it more than a little inconvenient to change to an alternative RTOS. The reason is the IDE, compilers, debuggers, third-party tools and idiosyncrasies are too weighty to change lightly. Engineers trained to one development environment will resist changing to another. An organization accustomed to one development environment, will tend to institutionalize that environment. In many situations, the decision to forsake an in-house OS and go commercial will carry collateral commitments that all but make further switching impossible. Imagine the implications of a last minute switch from GM's diesel injectors using Wind's tailored kernel to another OS. Absolutely inconceivable, as long as the kernel in question is adequate. As another example, INTS confessed that many pRISM+ sales have been sans their integrated tools (compilers, etc.) which reduced gross margins. Why would pRISM+ users choose to retain non-recommended tools? Because the pSOS developers are used to them and refuse to switch unless absolutely necessary. As a matter of fact, given the difficulty of getting many developers to switch from home-grown tools to commercial, I believe that barriers to switching in the embedded systems space are exceedingly high and rarely challenged.
HP is an important endorsement for WIND, but so is NT, NCI and many other major accounts. These endorsements and the industry's broadest range of tools and open development environment attracting third party developers are propelling WIND into being the only viable alternative to Microsoft.
As to the attraction of Microsoft, Ramsey wonders why industries would allow Microsoft to dominate their industry and extract usurious fees in the process. Actually, there is only one reason why Microsoft threatens to hold sway in embedded systems-a cheap, standard programming environment that is supported by tons of product developers and service providers. Exceedingly high run-time fees will be unacceptable, and this might cause Microsoft a huge headache. Bill Gates is used to charging $50 plus for run-time fees (no doubt causing the very problem Ramsey alluded to about percentage of costs associated with $500 PCs) which is completely out of range for most high-volume embedded systems. The desktop and server computer guys cringe when they discover that run-time royalties can be a fraction of desktop rates. Unless they bend to this level, they won't even be considered.
But don't forget that Microsoft has yet to run over the database guys, namely Oracle, and Microsoft has been in the database business for eons. They bought Foxpro in the 1980s, they brought out Access years ago, and they teamed with Sybase to build SQL Server so long ago that the divorce is long forgotten. As another example, Softimage was (is) a major 3D graphics program, running on SGI equipment and nipping on Alias' tails, when Microsoft bought the company years ago. Microsoft converted the program to work on Windows NT as well as SGI, and put their enormous weight behind the company. Today Alias is still number one with Softimage trailing, just like it was years ago.
The fact that Microsoft has been in the embedded systems business with Windows CE now for years, and just this week confessed that Windows CE lacks just about everything important in an embedded RTOS, should not give Microsoft embedded systems customers a warm and fuzzy feeling. Did they just discover these shortcomings? When they fix them, will the Win32 simplicity still be in place? Or, will the serious Windows CE be something completely different? Why did they agree to mix Windows CE with pSOS in order to get real-time and other missing features? This all sounds like Java's garbage collection stupidities. Windows CE is not ready for primetime. Microsoft knows it, developers know it and we know it. What none of us knows is when these problems will be resolved.
Now, what does it really mean if Microsoft is coming to the embedded systems party? Already INTS is willing to play the sycophant and add real-time features until sometime next year or so when Microsoft will have their own built-in features. The remnants of Microtec Research in Mentor Graphics are willing to replace VRTX with Windows CE and apply their XRAY debugger (which must be awfully long of tooth by now) to Microsoft's benefit. A number of systems integrators are constantly at Microsoft's beck and call to force-feed Windows CE wherever they can find an orifice.
Just by owning up to total inadequacy, Microsoft has begun a major shakeup of the industry. Unbelievable but apparently true. Nevertheless, the shakeup is good, irrespective of the dumb reason that caused it. What shake up am I talking about? Getting the message to bit players that they cannot endure with a business model built on run-time royalties. To succeed in this business, many of the companies that see themselves as RTOS vendors will need to revamp their image, and become applications consultants. They will need to change from providing cutesy toolsets to rolling up their shirtsleeves and helping product companies develop new products. The RTOS, IDE, compilers, etc. will be industry-standard supplied by WIND or, in the far future, Microsoft.
INTS has already signaled an eagerness to play by the new rules. So has Mentor Graphics. Microware needs to sign on, as well as most of the remainder of the RTOS holdouts (there will always be a few independents testing new waters, like Integrated Chipware). Third party library providers will luxuriate in being able to concentrate on just two RTOSs: VxWorks and a future Windows CE.
As big as Microsoft is, the embedded systems portion of Microsoft is relatively small, and caught up in a legacy design of Windows CE for palm held devices that is not up to snuff for embedded systems. WIND, backed by Intel (I2O), HP (JetSend and Java) and NCI (applications), among others, has much more capability, knowledge, focus and a huge head start. For every company that rejoices in turning its destiny over to Microsoft, at least two don't, and those are destined to be customers of WIND.
I can't honestly say that WIND will knock Microsoft out the embedded space over the next ten years, but I can say that WIND will at least establish one hell of a beachhead, worth billions in annual revenues.
If you feel Microsoft has had a disproportionate influence, relative to its contribution, on the embedded systems space, you are not alone. The Congress of the United States, the Department of Justice, huge players like Oracle/NCI, Sun, IBM and Intel are arrayed against Microsoft's dominance and their will will be done.
Allen |
