Feedback A forum for readers to speak their mind on issues that are important to the design community

Fear, but not loathing?

To the Editor:

I find your articles and the included opinions on the Linux versus NT debate to be typical--rooted in ignorance, fear, and bias.

Your table comparing what's supported in Linux as opposed to NT ["Linux vs. Windows NT: Engineers Speak Out, Part 2," August 1998, p. 26] contains many factual errors. Likewise, many of the people who speak out against NT show by their comments that they've never used it. For example, contrary to what your table says, NT users can readily find free gnu C and C++ compilers, Perl 5.0, and revision control systems.

Also contrary to your solicited opinions, not one Unix utility lacks a free or commercial NT counterpart. For example, the free Cygnus gnu C tools contain the complete set of Unix utilities (I myself prefer the commercial tools from MKS). There's also a free tcshell for NT that works extremely well.

Most engineers that I talk to about the issue compare a fully configured Unix machine to a bare-bones, out-of-the-box NT system. Does any commercial Unix box ship with tcshell or Perl? No--but all engineers assume that those tools are part of Unix. Well, guess what? Someone has to find and load the tools, whether on Unix or NT.

Here we run our engineering division entirely on NT machines--everything including Synopsys and Chronologic VCS. Our network was much easier to set up and maintain then a network of Suns would have been. It was cheaper, too: We all have 21-inch 1,600*1,200 true-color monitors and systems that cost one fifth of a basic similar-speed Sun (and the basic Sun came with only a 19-inch 256-color display system).

One engineer I talked to was vehemently opposed to our using NT. His number-one reason was that it was "unstable." When I probed further, it turned out that his only experience with a Microsoft OS was with Windows 3.1. I fear that most engineers who worry about the stability of NT have never used it. In one and a half years of operation here we haven't had even one crash (we've had to reboot, of course, when we've upgraded hard drives and performed other tasks).

As far as the ability to change the kernel source code, what kind of engineer interested in getting his job done is going to fiddle with the kernel source code? The people who think they do seem a little disconnected from the real world, in which engineers must actually deliver a product. Setting up an NT system is also many times easier and in the end far more productive in terms of the useful software available for free on the Internet.

I know that other people have had direct experience with NT as an engineering tool, and yet I never see them quoted in articles like yours. What possible purpose does the EDA press serve in knocking NT? Certainly the Unix companies are just as monopolistic as Microsoft (just less successful at it).

Pasquale Leone
Site manager
Zoran Toronto Lab
Toronto, Ont.

Murry Shohat replies:

Thanks for the comments. Clearly passionate about NT, Pasquale, you match the vigor found throughout my writeup of hundreds of readers' responses in the Linux versus NT series. As we observed in the articles themselves, the responses leaned heavily in favor of Linux.

You condemn the volunteered opinions of those ISD readers and the articles for being "rooted in ignorance, fear, and bias." To a large degree, though, ISD's treatment of readers' responses is similar to counting votes fresh out of the ballot box--but with one big difference: We were able to ask the voters, "Why?" By publishing the answers, we did our best to dispel ignorance. Moreover, no one at the magazine (myself included) has a stake on either side of the debate; we offered readers an opportunity to speak out and printed a representative selection of their comments. So your claim of "fear and bias" is a dog that won't hunt.

You are right about one thing: The table in my August article is incomplete as alleged. Before publication--but after the deadline--I received new information about Unix-like NT utilities including Cygnus gnu and the December 1998 release of GNUpro 98 r2. We've also received an even larger amount of fresh input about Linux utilities and full EDA applications, in addition to startling Linux support announcements from the likes of Intel, Sun, Avanti, and Summit Design. But in today's world any published technical data, even on-line, risks quickly becoming obsolete. So, though the table included immediately outdated information, I still think it serves a useful function by helping to bracket features and functionalities for objective comparisons.

Pasquale, your passion prompted me to reread my own work--all 12,000 words of it--and then digest your comments. Some are dead-on correct, as James Lee indicates in his response. Others, however, indicate that you're an NT bigot, just as two thirds or more of the readers who responded to our poll seem to be Linux bigots.

James Lee replies:

I'd like to draw your attention to the series of EDA PC-NT benchmarks we ran last year (we plan to run another series this year). The overall conclusion is similar to your experience: Simulation and synthesis can be productive and cost-effective on NT.

In the September issue I mention many of the same packages you do ("Linux Has What It Takes for EDA," p. 60). Certainly it takes more than the bare-bones OS to be ready to run EDA tools. I also agree with you on the kernel source issue. Although I have several Linux and NT machines, I've never needed to hack kernel source.

Nonetheless, once you've loaded the MKS or Cygnus kits, wouldn't you be better off on a native Unix system if the hardware were the same price?

Today, the best definition of an EDA operating system is the set of tools that run under it. If you want the whole ASIC EDA flow with the industry-leading tools, you'll still need your Unix machines--so don't throw them all away.

The not-so-free CPU

To the Editor:

As the guy who led the arch­design effort on the P6 (Pentium Pro, Pentium II, etc.) inside Intel, I got a kick out of your editorial on those folks who are donating their hardware expertise in an attempt to duplicate the software success of the Free Software Foundation ["Free the PC!" December, p. 6]. I know first-hand what they're up against in trying to design a leading-edge Intel-architecture product.

I bet this comment won't surprise you or them, then: I don't have very high confidence that they'll succeed. I do admire their chutzpah, though! I'd like to point out one other thing: The sheer coordination effort that it takes to get a large group of people to all move in generally the same direction is so huge--even when they all work at the same place and subject themselves to a common yearly performance­review process--that if the Freedom CPU folks can even turn out silicon that does something useful (never mind offering world-class performance), they will have succeeded at something of profound importance.

I mean this quite sincerely: If the project comes up with a way to usefully herd all the hardware "net cats" to a genuinely useful outcome, there would be real implications for all kinds of new endeavors. The team might discover a new and unanticipated positive use for the Internet itself: coordinating volunteer efforts to tackle important engineering tasks. Given who my employer is, I can't quite bring myself to say that I hope they succeed; but if they make even a good run at it, we could all learn very important lessons from the attempt.

Robert Colwell
Director of IA-32 architecture
Intel Corp.
Hillsboro, Ore.

To voice an opinion on this or any Integrated System Design article, please email your message to miker@isdmag.com.

integrated system design  March 1999