jep.jpg (13389 bytes)

Chaos Manor Special Reports

WINDOWS 2000

Monday, June 18, 2001

Email Jerry

Sections

Chaos Manor Home

View From Chaos Manor

Reader Mail

Alt.Mail

Columns

Special Reports

Book &; Movie Reviews

Picture Gallery

Links

Table of Contents

What's New

The BYTE Fiasco

A DISCUSSION

 

 

 

 


I had a great deal of mail referring to the Big Bug Report, which you can read at

http://www.zdnet.com/anchordesk/story/story_4466.html

and I suggest you do so. The answer to that one is simple enough: they ran a bugfinder on the source. This is a bit like running a grammar checker on a Hemingway novel. You will find a lot wrong. Whether that's real bugs or not is something else. I haven't had any informative letters on that subject yet. Most of the mail has been from Mr. Erwin.

The value reported is 63000 in 35 million lines of code. That is between 2x and 5x (depending on how you estimate it) the rate the Government is willing to tolerate when accepting a software system. (See page 118 of Musa, et al., Software Reliability, for some comparison figures.)

Harry Erwin

When I replied:

>Come now, that's the number found by a program similar to Grammatik. Run >Grammatik against Hemingway and see how much it finds and how much of that >you would change.

Jerry, they run similar programs against anything built for the Government, and anything that fails gets written up. I'm assuming they've got 63K open problem trouble reports. I've also heard that there are only 27K real problems. That's a reasonable fraction based on my experience with air traffic control systems, but it's still a high rate. Generally, a system being deployed is expected to have less than one open PTR per KSLOC of development effort, and those are supposed to be no worse than cat 3 (minor with a workaround available). Note that the Government is planning to use Windows 2000 for critical air traffic control applications.

 

 


Then there is the speed issue:

Jerry,

"Pat Gelsinger, vice president and general manager of Intel's desktop products group, said in a speech that Windows 2000 requires up to 250 more megahertz of chip power for performance equivalent to Windows 98 or the corporate Windows NT. That seemed to be a fairly large--and surprising--speed boost, according to analysts at the Intel Developer Forum here."

Full story at:

http://news.cnet.com/news/0-1003-200-1551163.html

Now, I know that it's supposed to be a *privilege* to keep shelling out more money for yet faster chips so we can all run the very latest Redmond offering at full waddle. But, to my way of thinking, this really makes Linux or LinuxPPC look better and better. Linux has done nothing but get more serious in the last year: more support for more devices than ever, a good solid modern kernel with the bugs well shaken out (2.2.14), and filesystem journaling in the works. Meanwhile, LinuxPPC has just come out with a built-in MacOS emulator. That's right now.

The near future? When enhancements for the gcc compiler come out which take full advantage of the G4's Altivec processor, LinuxPPC will essentially become the first 128-bit operating system. At 400 MHz for the $1.6K Powermac, that's going to be *quite* a bit of affordable bit-crunching.

And if VMWare (http://www.vmware.com/products/forlinux.html) comes out with a LinuxPPC port (they already run solidly on Linux for x86 chips), the whole situation is going to get utterly ridiculous: LinuxPPC will have 128-bit capacity and effective MacOS emulation *and* effective Win98 emulation in *one* multitasking system.

And even if LinuxPPC doesn't immediately have Altivec optimization, MacOS X is likely to -- unless Steve Jobs suddenly becomes dumber than he's shown signs of being for the past three years. MacOS X will offer slightly less flexibility but may well be the first UNIX system to actually be somewhat usable by normal non-wizards.

If you get the impression that I think Microsoft is in far more trouble than it dreams, you are right.

--Erich Schwarz / schwarz@cubsps.bio.columbia.edu

I can only reply that I have been using Windows 2000 for months on the same machine I used for NT 4, namely Princess, a Dual Pentium 200 Compaq Professional Workstation. I believe it is actually faster: I no longer have to run programs to find and eliminate memory leaks. It is certainly more stable: I have not had to reset in a long time, and every time I did it was due to some problems with Office 2000/Outlook 2000 which can be jammed by large and complex mail and web sites when accessing over a fairly slow connection. In at least two cases I was able to duplicate the "crash" and determine that had I waited about 70 seconds I would have no problems. NT has at least that many delays with Outlook problems, plus others.

On stability 2000 is way ahead. On installation it's wonderful: most of the plug and play actually works.

But understand I mean only Professional = Workstation. I have no real data on Windows 2000 Server. I have little hesitation in recommending 2000 Professional, always with the caution, if it ain't broke don't fix it: there's no point in upgrading to 2000 if your NT system is satisfactory, you don't need USB, and you aren't contemplating major hardware upgrades. With Windows 98 it's a bit of a different story, but that's for the next column.