Dr. Dobb's Journal July 1998
More than once, Microsoft's marketing knuckles have been rapped for its nasty little habit of preannouncing products. Not that rebukes seem to matter, if Microsoft's recent preannouncement for adding "hard" real-time capabilities to Windows CE is any indication. According to a somewhat nebulous press release, Windows CE will be a "hard" real-time operating system with the release of Version 3.0 sometime in the second quarter of 1999. Interestingly, at last fall's Embedded Systems West Conference, Microsoft was careful not to call Windows CE 2.0 "hard" real time at all. Then suddenly, like pigs sprouting wings, Microsoft was referring to WinCE as a hard RTOS at this spring's Windows CE Developers Conference. (Soft real time is more forgiving than hard. Soft real time can miss deadlines in cases where not completing tasks is more acceptable than a failure. Hard real-time deadlines, on the other hand, must always be met. At minimum, hard real time must be deterministic, have low latency, and support nested INTs.)
When asked about this in a DDJ web site "Online OP-ED" interview (http://www.ddj.com/), a WinCE product manager cleared things up, explaining that the 1999 release of WinCE 3.0 will be "true" hard real time, implying that WinCE 2.0 is some other kind of hard real time. Yes, by the most minimal of definitions, WinCE 2.0 is a RTOS -- but it's about as hard as butter on your morning biscuits.
WinCE has a long way to go before it can truly be called hard real time -- especially when compared to tried and tested RTOSs such as QNX, VRTX, VxWorks, pSOS, and the like. (For instance, some WinCE latency figures are measured at from 93-275 microseconds; under QNX, comparable figures are at about two microseconds.) In all likelihood, a total rewrite of the WinCE kernel will be required to bring WinCE up to par with real RTOSs. But for all we know, of course, that rewrite is underway.
If history has taught us anything about Microsoft, it is that the company has a hard time meeting promises when it comes to shipping operating-system products -- especially when those products are announced more than a year in advance. Having more resources than most of us can imagine didn't necessarily get Windows 95/98/NT 5.0 out the door when promised, making you wonder why Windows CE should be any different.
So why does Microsoft keep on preannouncing operating-system products so far in advance? More than likely to freeze the marketplace until a minimal implementation of what's promised can be delivered. Clearly, that marketplace would be better served by walking the walk, instead of talking the talk.
***
About the same time Microsoft was exhibiting chutzpah in the real-time realm, O'Reilly & Associates was down the road patting itself on the back over its self-proclaimed "historic" Open Source Summit. According to its press releases, O'Reilly brought together "heavyweights of the Internet software community...to explore ways of expanding the use and acceptance of open source software development."
No question, this is an admirable goal that deserves all of our support. The invitation-only event included the likes of Linus Torvalds, Larry Wall, Brian Behlendorf, John Ousterhout, Guido van Rossum, Phil Zimmermann, John Gilmore, Eric Raymond, Tom Paquin, Jamie Zawinski, Sameer Parekh, Eric Allman, Greg Olson, and Paul Vixie -- each of whom deserves accolades for his contribution to the world of software development.
More noticeable, however, was who wasn't invited. If any single person deserves credit for launching the open source software movement, it's Richard Stallman of GNU and free software fame. An "open source summit" without Stallman is like a cheeseburger without the cheddar.
When, in response to a flurry of O'Reilly e-mail press releases, I asked by reply why Stallman wasn't invited, the net went suddenly quiet. Inquiring minds want to know.
***
A recent study by Software Success (http://www.softwaresuccess.com/) revealed a couple of interesting twists. The analysis, compiled by Software Success using data supplied by Dun & Bradstreet, showed that the total number of companies competing in the software industry grew from 58,779 in July 1997 to 68,765 in March 1998. For the first time since 1993 (when Software Success started tracking this data), the rate of growth of mid-sized companies was faster than that of startups. For instance, the number of companies with annual sales of under $500,000 (50,482) increased 12 percent since July 1997, the number with sales of $1 million-$5 million increased 42 percent, and those with $10M and up increased 83 percent. Software Success also found that the number of companies in the software-related services sector grew 45 percent to 19,542, reflecting a migration of some formerly product-based companies to the services sector.
--Jon Erickson