Sometimes debugging software (or a process) is like playing with those Russian nesting dolls: you crack one problem only to find another one deeper inside. Fortunately, they tend to get smaller as you go along. 8-)

Two weeks ago, a relative handful of subscribers got two copies of the last LangaList. Turns out there was an ISP in the Netherlands (the LangaList goes all over the world) whose mail server was set up incorrectly. That mail server was bouncing the List back to my listserve in a way that caused it to be sent out again.

Fortunately, we caught it as it was happening and not too many of you got duplicates. I wrote to the subscriber in the Netherlands whose account caused the problem, and also notified the webmaster of his ISP that the mail server there was broken or misconfigured. I then blocked all further mail from that ISP, and changed the settings on the list serve so that even if some other mail server does the same thing in the future, no duplicates should result.

The above took a while, which is one reason there was no issue last week. The other is that I was celebrating my son's birthday. We've developed a tradition where we spend weeks planning a father/son adventure, and then spend a day or a weekend off by ourselves doing something silly and (I hope) memorable.

Through it, we found a park in Massachusetts that has the second-steepest coaster in the world--- a mind-numbing 58 degree drop, which immediately twists into about a 60 degree banked turn.

It's an old wooden coaster so it creaks and groans; the whole structure vibrates, sways and shakes, and you can feel the wooden cars flexing beneath your butt as the car careens along the track. It's a scary sucker!

There are several other coasters there including a modern inverted steel number and a replica of the wooden coaster from the 1939 World's Fair. (We rode them all.) But the old wooden "Cyclone" is the main show. It was great!

So, instead of writing a List last weekend, I was off wreaking havoc with my inner ears and sense of balance---and I hope building a nice lifetime memory for my son.

OK, back to business:

If you're running Office 97, I have good news and bad news. The good news is that there's a new pack of bug fixes, updates and improvements to Office. The bad news is that it's a 23 MB download!

It really is free---no "shipping and handling" add-ons and such. Microsoft really is playing nice with this patch, to their credit. Thanks, Bill!

There are a few legal and technical gotchas. The free CD offer is available only to US and Canadian residents. And both the CD and download versions require that you enter a valid Product ID number before they'll work.

But perhaps the largest speedbump is that before installing SR2, you need to be running Service Release 1, a 7MB patch that came out shortly after O97 originally shipped.

You can tell what you're running by opening any Office application's Help/About box and look to see if "SR-1" appears in the name. If you run Microsoft Word, for instance, you'll see the product identified as "Microsoft Word 97 SR-1" if you're running SR1. If not, you'll need to download that patch (also free) and install it before installing SR2.

Maybe it was wishful thinking, but I really thought Microsoft had solved this problem. I certainly had hoped they'd solved it. I was wrong.

Win98 is supposed to keep track of critical system files and manage conflicts so that newer files don't get overwritten by older files as you install new software and drivers. The two front ends to the system are the "System File Checker" (SFC.EXE in your WINDOWS\SYSTEM directory) and the Version Conflict Manager (VCMUI.EXE in the WINDOWS directory).

But they don't work very well, and I'll bet you lunch that right now, your system has mismatched DLLs or is running older files in place of newer ones. I'll tell you how to check your system in a minute, but first let's see some examples from my own systems.

Here's how it's supposed to work:

Somehow, even on my clean-install version of Win98, I ended up with two copies of JOBEXE.DLL, which is part of the Windows 98/Internet Explorer Active Setup library. Win98 had originally installed version number 4.x.3 on my system, but somewhere along the line---probably when I ran Windows Update---I ended up with a slightly newer version, 4.x.7. WIN98 correctly noticed the difference in version numbers and installed the newer file, making a backup of the older version. Nice!

So far, so good. But I also ended up with two versions of ODBCJT32.DLL, Microsoft's ODBC Desktop Driver. After comparing the version numbers, Win98 backed up version 3.40.x and installed 3.50.x And that looks correct, doesn't it?

Trouble is, the "3.5" version is smaller and a year and a half older than the "3.4" version. My guess is that some vendor along the way played with the versioning: Win98 relied solely on the version number and was tricked into installing the much older file, thinking it was newer.

Win98 can be tricked other ways, too. For example, I found files carrying the exact same version number but that were different in size and created on dates as much as two years apart! But because the version numbers were the same, Win98 treated them as equals, and as often as not installed the older version.

There are many other examples---I was dismayed to discover I actually had several dozen instances of older system files replacing newer ones.

I just found this out late last week, so I'm still sorting through what it means and what to do about it. My guess, though, is that this may be one major reason why some installations of Win98 are unstable: Win98 thinks it has the latest version of some key files, but in fact may be running obsolete system components from the early days of Win95.

Want to check your system? Run SFC.EXE, and---this is the key thing---click on the "Settings" tab. At the bottom of the Settings page, check off both "Check for changed files" and "Check for deleted files" and click OK. Now (on the main page) click "Scan for altered files," and click Start. I'm betting you'll be very, very surprised at the number of possible conflicts you find.

Next, try VCMUI.EXE, and there's a good chance you'll find similar results there, especially if your Win98 installation is an upgrade rather than a clean install.

You've heard about the "Year 2000" problems: some hardware and software incorrectly treats the year 2000 as the year 1900 or 1980; other systems mishandle the leap year that year, and so on.

Late last week, the San Jose Mercury News broke a small Y2K-type story I hadn't seen written about before: It turns out that when the Government designed the Global positioning system some 20 years ago, memory constraints caused them to run the ultra-precise GPS clocks on a one-kiloweek (1024 weeks) cycle.

Guess what? The kiloweek cycle is almost done. In exactly a year---at 17:59:00 EDT on Aug. 21, 1999---the GPS chronometers roll over and begin a new 1024-week count, starting from zero.

Most new GPS units won't be bothered by the kiloweek rollover, but some older models suddenly could freeze or produce wildly inaccurate results. You probably wouldn't want to be far at sea in a ship, or high in the clouds in a plane with one of these old units next summer.

Worse, many of those same older units may also have trouble just a few months later with the classic Y2K bug itself.

The SJMN report was refreshing because it was so specific---in marked contrast to many of the Y2K reports cropping up. The last issue of Wired, for example, featured interviews with Y2K techno-catastrophists who have left their jobs as programmers or IT professionals and literally headed for the hills: They truly believe civilization will grind to a halt on January 1^st, 2000 as the electric grid, communications systems, transportation network and banking infrastructures grind to a halt. As a result, they're holing up in remote shelters and boning up on obscure skills such as tanning hides and subsistence farming.

They even have a new piece of jargon for the Y2K effect: TEOTWAWKI. It's pronounced TEE-AWT-WAW-KEY, and it's an acronym for "the end of the world as we know it." I'm not kidding: They are dead serious.

Meanwhile, until Wednesday, last week's CMPnet column has sparked a firestorm of exchanges. In that column, I wrote how Microsoft is between the proverbial rock and hard place, and how problems will soon deepen as the courtroom phase begins in several weeks. There are many issues involved, but the one I focused on is one that literally affects what you're doing every time you go online. It's about your browser.

What's your browser of choice, and why? If you're using IE, how much of your choice was influenced by the ease with which you got your copy (i.e. bundled with Windows, for example). If Communicator were on the Windows CD, would you use it instead?

If you're using Communicator, how much of it is a desire not to support or use a Microsoft product? If your choice is based on technical merits, what are they? What else made you choose Netscape over Microsoft?

Another record month at BrowserTune (http://www.browsertune.com)! I've continued making numerous small enhancements to BrowserTune to fine-hone as new sub-versions of the browsers come out, and to reflect the new-found security issues I've been telling you about in past issues of the LangaList.

This last couple weeks, I've gotten more emails from users whose copy of IE4 has lost the ability to follow links; and who are getting garbled text on IE4 pages. Both of these are definitely IE4 bugs, and the "Browser Failed?" pages at BrowserTune contain suggested fixes.

A new version of Communicator is out, too, and I'll be putting that through BrowserTune soon.

The HotSpots page (http://www.browsertune.com/flanga/hotspots.htm) remains the single most popular page at browsertune.com; it's wonderfully gratifying to see how many people visit that page every day.

Two weeks ago, I asked for volunteers to assist in pre-screening the flood of thousands and thousands of hotspot nominees I get each month. Several people volunteered (thank you!), and we're working out the details now. Once we have it down pat, I'll list the names of the volunteers here and on the HotSpots page.

But here's a curious factoid: almost all the volunteers were women. Given that most WinMag readers are male, and that most people on the web in general are male, having an almost-totally female response was surprising.

Anyone care to draw a conclusion from this? I'm not even going to try--- except to sincerely thank all those who graciously volunteered to help! 8-)

I've been trying to include a short silly item at the end of each LangaList. The last couple ones have brought some email such as this:

Fred

(fred@langa.com)

An easier-to read formatted HTML version is available in the "what's new" section of http://www.langa.com. All past LangaList issues are also available via the same link.