n do at your end to make things faster; the Internet is getting jammed up, and popular sites are experiencing server overload.
Sometimes, though, the problem can be your serial ports. Older computers may not have high-speed serial-port chips. That will really limit communications. Newer ones (and all new internal modems) have the 16550 "high-speed" universal asynchronous receiver/transmitter (UART) chip, which is fast enough for modems on normal telephone lines, but too slow to get the full benefit of ISDN. When you get above 80 Kbps, the ports often can't keep up.
The simplest remedy for this is Lava Computer's Lava Link-650. The Lava Link-650 has a 32-byte buffer (as opposed to 16 bytes in the 16550) and enough smarts to recognize flow control at the chip level rather than working your CPU. If you normally do communications in the background while running other tasks, this could be important.
You can install the Lava Link-650 as any of the COM1
to COM4 ports, and it accepts any of the normally unused interrupt requests (IRQs) between 2 and 15. It works under Windows 3.1 or 3.11, Windows NT 3.51, and Windows 95 (Win 95). It does not work with DOS. You can run DOS communications programs under Windows just fine, but if you do that, you're not likely to need a faster serial-port board anyway. I haven't tried the Lava Link-650 with OS/2. Lava is working on a native OS/2 driver, which may be ready by the time you read this.
I've yet to make ISDN do anything useful, but I haven't worked very hard at it. I'll get to it Real Soon Now, and when I do, I'll run it through the Lava Link-650. ISDN is 128 Kbps, and the best your serial port can do is 115 Kbps; without something like the Lava Link-650, you will probably lose data, and you'll certainly get retries and resends. It will be interesting to see if ISDN and faster ports bring about any real improvement in Internet connect speeds. I suspect they won't, but I'm prepared to be pleasantly surprised.
There was a time when I routinely tried new stuff under OS/2
as well as Windows, and I kept this up even after Win 95 came out. I'm open to argument from readers, but I've pretty well decided not to do it anymore.
It's always hard to understand what IBM's plans are, but it looks to me as if despite the investment in the Merlin upgrade, IBM is abandoning OS/2 Warp at the consumer level. They'll continue to support it for corporate clients who want it, but I don't think they're going to be very aggressive about pushing OS/2 to the public. Instead, they're working on making all their other software, particularly server software, compatible with just about everything: DOS, Windows 3.x, 95, NT, Unix, and anything else they can think of, and tying it into their excellent database products.
Given IBM's track record, this may be the right thing to do: the company has some of the brightest technical people in the world, but it's hard to understand where they get their executives. In the old m
ainframe days (like in 1964, when they bet the company on the System/360), IBM's marketing was literally fabulous, but in the consumer market, they can't sell eternal youth at an AARP meeting. Incidentally, when Comdex opened in Chicago, we saw a sign designating a large suite of rooms as the "IBM Media Relations Control Room." Several BYTE editors, including me, stopped by to be controlled, or at least to get a cup of coffee and find out about Merlin, but we were told this wasn't the place for the media to go; this was for IBM employees only. By the third day of the show, the sign had been changed, and I'm told that they were welcoming press people, but by then I had no time.
Merlin, the new release of OS/2,
has IBM's nifty voice-recognition software built into the OS. That's good, but the software is also available for Win 95 systems, and besides, most people who work a lot with voice recognition prefer Dragon Systems' newest version. Another case of a great idea but late, late, late.
Merlin will not take advantage of dual processors. Neither, at the moment, does any of the OS/2 server software except for old, vanilla, pre-Warp OS/2 2.1. Meanwhile, off in the back room we have Spirit, a dual-Pentium 133 Micronix motherboard system running NT and acting as the Chaos Manor server. It has been running for well over a month without any shutdown or reboot. I haven't done a lot with it because I haven't had to: it just works, splendidly.
We did install the big Exabyte 8700 8-mm external tape drive.
Our first attempt failed due to cables, but a new Granite Digital SCSIVue Diagnostic Gold Cable fixed that. Pournelle's law: if you're installing SCSI, start with a Granite Digital cable.
We hadn't really missed having a backup system because Spirit is so fast and has such humongous disk space that I can use my famous "multiple-redundancy" backup system: I save everything to both the Digital Equipment 3-GB internal SCSI hard drive and the Micropolis 4-GB external drive, as
well as keeping a copy on the machine where I created the file. I know this isn't efficient, but these days, it's so fast I don't care.
NT is an easy server system to use
because to the rest of my network, it looks just like another Windows machine. That's not quite true: my Windows machines see Spirit's disk drives and files nicely, but if I ask for "properties" on one of Spirit's networked drives, I'm told it has 2 GB, all empty, on each one. To find out the real status of those drives, I have to go over to the Spirit machine's console. This is due to using the NT high-performance disk format, and it's hardly onerous.
Meanwhile, there's bad news: the upcoming revisions of NT will change the way the OS handles the graphical device interface (GDI). Previous versions of NT left the GDI out in ring 3, or user mode. The new version of NT treats the GDI as Win 95 does, by giving it ring 0, or kernel-mode status. This change comes to boost NT's performance as a client OS, but it has the po
tential defect that if, say, a graphics driver croaks, the whole system could crash. Drivers can be corked by bad applications, including programs that violate the cynical version of the first rule of system programming: "Never test for an error condition you can't handle." Of course, the real rule should be, "Test for errors and be sure to have a way to handle them when you find them."
The result of this will be to make NT a bit less robust in new versions and a bit more vulnerable to bad applications. It will have one more failure mechanism on mission-critical applications. I realize that many people think this will turn out to be a non-issue, but if this vulnerability proves real, it will be a blooming shame since many people bought off on NT for crash-proofing.
Meanwhile, although the situation is better than with Windows 3.1, really bad applications can so bollix up Win 95 that there's nothing for it but to reinstall from scratch. If that happens, you probably should go looking for two hidden sys
tems files -- user.dat and system.dat -- and kill them dead before you reinstall. Those files are modified but not replaced when you reinstall Win 95 over an older installation, and sometimes those files are causing the problem that made you reinstall in the first place.
The safest procedure is to go in with fire and sword, kill those files, and rebuild the village stick by stick.
Incidentally, old Norton Commander running in a DOS window can see those files and delete them. There's never been a handier system utility than Commander, and until someone makes a Win 95 version, it will be one of the very first things I install on any Windows or Win 95 machine I use.
I have mixed emotions about CyberMedia's FirstAid 95 Deluxe.
It seems to work quite well, and it's about the best program I know for analyzing your Win 95 system and finding dead-end shortcuts, useless DLLs never called by anything, and all the other garbage that seems to accumulate in Win 95.
It runs in the backgroun
d to intercept errors that might cause a system crash, and I suppose it does that. I've certainly seen it tell me about application program errors, which it says would have brought the system down if it hadn't caught them. I ran the same bad application on the Cyrix system and was able to recover through Ctrl-Alt-Del and turning off the nonresponding application. The FirstAid 95 method is a lot more convenient.
There's also a new feature that will automatically dial up the FirstAid database to see if there's any new information about your problem. It will, on command, automatically go on-line to the Internet to get the latest Windows Service Pack (Microsoft's name for what IBM calls a Fixpack) and apply it. These are potentially wonderful features, and I don't know of any other program quite like this.
In a word I like FirstAid 95 Deluxe, but on the Cyrix system, and I suspect on any Pentium Pro system, you simply cannot run FirstAid 95 Deluxe and Norton System Doctor together. I couldn't even get the
Cyrix system to boot up after I installed FirstAid, which, incidentally, automatically puts itself into the Startup folder. I had to boot up the Cyrix system in safe mode -- hold down the Shift key while the machine boots -- and drag the FirstAid program icons out of Startup and onto the desktop before it would come up. This was surprising, because I'd had both FirstAid and System Doctor running simultaneously on Pentafluge for weeks without a conflict.
Well, all right: there's no real need to run both FirstAid and System Doctor. Choose one. For the moment, I'm more comfortable with System Doctor, and anyway I'm still doing tests.
And that's where my objection to FirstAid comes in. After I dragged its icons out on the desk, I looked into win.ini and system.ini to see if there were any instructions for loading or running FirstAid. None. But when I rebooted, there was the little FirstAid icon down in the lower corner of my taskbar, and Ctrl-Alt-Del showed several of its programs were running. Worse, th
ey were taking up a big block of CPU time, actually slowing the system down. I took another look for traces of FirstAid in init files, found none, shut down, and this time I turned the machine off for 10 seconds before restarting. When it came back up, there was the FirstAid icon in the taskbar. It's red, to show it's not running, but it's there, and I have not the foggiest notion why or how to get rid of it. I suspect I will have to uninstall FirstAid with fire and sword, possibly using CleanSweep or some other cleanup program.
So. FirstAid has done well for us in the past. Their cleanup program is as good as any I know. The dial-up features are potentially extremely valuable. It works automatically. It seems to behave well on Pentafluge. You can turn it off. On the Cyrix system, it will not run simultaneously with Norton System Doctor, but that's not important. My big objection is that once FirstAid 95 Deluxe gets into your system, it doesn't want to go away, and that scares me. More when I know more.
Do you like mushpads? Well, the technical name is touchpad;
these are the mouse substitutes that have a touchscreen surface. You move your finger across them to move the cursor. Mushpads have long been standard on Mac PowerBooks, and more and more laptops now have them.
The problem in the past has been click and drag. Cirque's GlidePoint Touchpad 2 has software that fixes that. You can click by tapping the pad, click and drag by tapping once and then again, but don't lift your finger; you can now drag. Double-click is of course two taps, removing the finger both times. If this sounds complicated, it isn't. You can learn it all in a few minutes and get thoroughly used to it in an hour.
I'd far rather have a mushpad than a "J key" or one of those other eraserhead pointer things. My Gateway 2000 Liberty laptop has an eraserhead pointing device, and I sure wish there were some way I could replace it with the GlidePoint. I don't wish it quite so much that I'm willing to put in the Glide
Point Touchpad 2 as an external device attached to the serial port, but it's close.
The new software also solves the problem of running off the edge of the mushpad; that used to drive solitaire players bonkers, but it's fixed now. First, the whole pad is larger. Second, there's an edge-track area; drag into that, and when you let go, it's like lifting the mouse to get more rolling room.
In fact, everything is fixed, and the GlidePoint Touchpad 2 is at least as easy to use as a trackball; possibly as easy to use as a mouse. I'm pretty used to the Microsoft "Big Teardrop" Mouse 2.0, particularly for playing games, so it's questionable whether I'll switch to mushpads on my main machines; but I have to say I'm slightly tempted. The GlidePoint Touchpad 2 looks good, it works, and it's a sure conversation starter. Since you can rest your arm/wrist on the desk while using it, there's less strain on the wrist and hand. I don't particularly notice wrist strain from mice except at dawn after an all-night sessio
n with War Craft II -- but I do notice it then, and this may be the answer.
Cirque's mushpads are rugged; we had one on Little Cheetah, a general-purpose machine in the Great Hall, for over a year, and it was used to play Railroad Tycoon, Master of Orion, and several other games, as well as for real work, including word processing. It was dropped on the floor, had junk stacked on it, and was in general mistreated, and it still works. The new models look at least as well built.
If you're weary of mice, try Cirque's GlidePoint Touchpad 2. You might like it. They also make a keyboard with a built-in mushpad, called the GlidePoint Wave Keyboard. That looks interesting, and I'll try one Real Soon Now. My problem with keyboards is that I love my old Northgates in the "Pournelle" configuration -- the Backspace key is on the
yuiop
row where God intended it to be, rather than up next to the numbers -- and I have no temptation whatever to change. Those who don't mind the dumb key layout that seems to h
ave become standard might want to look into Cirque's GlidePoint Wave Keyboard.
Product Information
FirstAid 95 Deluxe......................$59.95 (estimated)
CyberMedia
Santa Monica, CA
Phone: (800) 721-7824 or (310) 581-4700
Fax: (310) 581-4736
Internet:
http://www.cybermedia.com