Why I moved most of my PC work back to DOS. If I knew of a simple, multitasking operating system with good application support, I'd change.
Most of the notes herein were taken between when I unpacked new parts for my computer in late August 1998, and when I stuffed this on a web site early 1999. I don't spend my entire life using a computer, and I had lots of other stuff to do in this time, so it wasn't like I'm a full time computer user. A lot of computer systems seem aimed at full time users, and for all I know, are a wonderful match to such people. However I want to get my jobs done with an absolute minimum time using the computer, and a minimum of overhead outside the current work.
I'll add that I'm not entirely unaware of how computers in general operate, although I've usually avoided PC hardware. I've built several 8 and 16 bit computers in the distant past from kits or from bare boards (by which I mean bought all the components, soldered them in, made up all the cables and cases, etc.) I've written drivers and little utilities in Z80, 6502 and 68000 assembler, and used a few different compiled languages. But PCs and Wintel are pissing me off more than anything else I've ever attempted to use. I say attempted, because the support overheads on the PCs I've tried have been ridiculous. I've lost count how much effort I've wasted stuffing around instead of working.
At first I seemed to have no alternatives to DOS and or Windows 95 if I want to run an Intel computer, since many other operating systems don't always work on the Wintel box I had available. Since the Intel boxes are cheap, using them seems otherwise desirable.
Pick an Operating System
BeOS Release 3.0.1 for Intel #70012465. The boot floppy loads, puts up a message about booting, and then allows only one option. "Rescan for boootable volume". Fails to find my (standard IDE) CD-ROM. Can't boot direct from its CD-ROM, although my BIOS allows that option. According to the BeOS hardware list, my PC has the correct components (naturally I checked the list before I bought BeOS).
That really did disappoint me, because it really sounded as if BeOS had the cleanest OS design around. I gather BeOS are getting out of the PC OS business, although I've heard they make the later versions (v5?) available on their web site (April 2000).
Caldera Open Linux Lite does better, although you could do with a manual before starting it (I had a free copy handed out at Comdex 1997, so not surprisingly, there is no manual). Once you figure out the partitioning (code number for the swap partition mainly), and recall how to mount the CD-ROM (not intuitive if you haven't used Unix in a while), the command line interface works. However the X Server fails to create the XF86Config file when you run XF86Setup. Manually copied video driver to /usr/X11R6/XF86_S3V (which was why I wanted the CD mounted). XF86Setup runs, but fails to communicate with the X Server. I never did solve it, so I never did evaluate this Linux.
IBM's OS/2 Warp 4 installs just fine, however once I started looking at getting everything working, it was less impressive. No driver for my STB S3 Nitro 3D Virge video card. The OS/2 v3 driver doesn't work. Nothing useful on the OS/2 Driver CD. Best resolution I can get is 800x600 in 16 colours, which is useless. The Creative Labs Sound Blaster 16 sound card seems to be detected, but doesn't actually produce any sounds. The Eiger PCMCIA card adaptor isn't supported at all. The Hewlett Packard LaserJet 5MP printer doesn't work (although interestingly it still works fine under IBM's version of DOS - I'm guessing it wants a fancier printer cable than I have). OS/2 did support my Trantor T130B SCSI hard drive adaptor (which Windows does not).
So I had to try Windows 95 (which at least had the merit of loading, and supporting my video card and sound card), but it doesn't support my Trantor SCSI controller, so I can't get at my data drive. I finally got my Adaptec 1520 SCSI controller back, and that works fine under DOS, but only if the SCSI drives are below F: Under Windows 95, it reports a conflict with the video card (there isn't - the conflict is with the Creative Labs sound card). Windows also doesn't support my Eiger Labs PCMCIA adaptor, so I can't read flash rom cards or sram cards. I found I could install DOS drivers and access the SRAM cards at least - it seems neither Windows 95 nor Windows 98 support SRAM PCMCIA (the original reason for having PCMCIA). Wow, I'm impressed.
Windows 98 would not complete installing, and would crash without getting very far into the install process.
Windows NT version 4 would not complete installing, and would crash without getting very far into the install process.
Red Hat Linux 5.1 installed without any problem. Handled all my drives, including those on the Adaptex 1520 SCSI controller. It worked great with my PCMCIA adaptor, on SRAM cards and Compact Flash. X Windows installed fine. About the only things that I couldn't figure how to use were my WinRadio ISA card (but there is now a beta driver), and my parallel port scanner (I couldn't find a SANE driver for it). Of course, the PsiWin palmtop software isn't available, but that only exists for Windows 95/98. When I have some time, I rather suspect one of the updates of Red Hat might be in my computing future.
Had some floppy disks with "backup.001" and "control.001" files on them (from a third party). Took an MS-DOS prompt and tried to run "Restore". "Wrong version number". Figures!
Booted into MS-DOS 5 instead of Windows. Restore couldn't recognise the backup files. Found a machine with MS-DOS 6.22 and tried the same. It doesn't recognise the backup files.
Back to Windows 95, and used "Start", "Help" and selected help that said "Backup, starting". "Click here to start backup". Fails with a DOS box saying you didn't start it right.
Went via "Start", "Programs", "Accessories", "System Tools", "Backup". Backup starts, tried the "restore" tab. Doesn't even detect the files.
Why in the hell does Microsoft even bother to provide this? I've never yet seen it work correctly from one machine to another.
Radsoft.net report on RegClean, and how a 153,600 byte executable could be cut down to 69,632 bytes in ten minutes. So why didn't Microsoft do just that?
At the same site, the Bloatbusters report on the Solar Winds DNS Resolver. The download is 3,666,493 bytes. Bloatbusters wrote a replacement, with the same functions, in under ten kilobytes. The Solar Winds product was written in Visual BASIC with heaps of OCXs, instead of with Berkeley Sockets API.
Got a new motherboard to replace the one with the dead primary IDE controller. Turns out that it won't tolerate the IDE CD-Rom as a Primary Slave drive. Moved it to Secondary Slave. The already installed Windows 95 won't detect the CD-Rom (although the BIOS sees it). Can't convince New Hardware to detect it. It does detect some gadget and asks for a file ... from the CD-Rom that it can't read. No copies of the file on any other PC. Tried to copy it off Windows 95 floppies ... nope, doesn't exist on them. Finally used PowerZip to extract it from the Window 95 CD using a different PC. However that doesn't get the CD working. Gave up and did a config.sys and autoexec.bat and included the CD drivers in DOS mode!
A little later we upgraded to Windows 98. That did support the CD, and even removed the mscdex.exe from the autoexec. Didn't catch the CD drivers in the config.sys, but that might be asking a little much.
Cookies and Web Bugs - nasty
Cookies are pretty harmless, but check this idea. http://www.tiac.net/users/smiths/privacy/cookieleak.htm
Copy from CD
Tried to copy an executable file from a CD to my hard disk. Explorer didn't give any indication it did it. Tried right clicking on the file and selecting copy. Still no result. Noticed after a few tries that I now had about three "Shortcut to file.exe" on my hard drive. Tried dragging and dropping with the shift key down, with the Ctrl key down. Now I have five shortcuts. Lost my temper and copied the thing in DOS.
Not mine, but my partner's drive. And when we plugged in the alternate drive, it started to give problems (which made me believe we have a motherboard IDE problem). Installed the drive in my PC. Both copies of the FAT are a mess, the root directory hasn't any subdirectory showing. Tried using Norton (from another drive) to do a fix, but the automatic repair left the drive unreadable. The only files I could recover were done by scanning physical sectors and extracting file fragments (you know how much time that takes!) And that was done using the DOS versions of Norton, since that never seems to have been updated. I was royally pissed of by the time I'd recovered the few files that weren't saved elsewhere.
DLLs, or Dynamic Link Libraries, are reusable software functions or components. Applications can share DLLs that are supplied with Windows. Unfortunately, Microsoft chose to distribute Windows 98 with some DLLs that are older than the versions supplied with Windows 95 version C. Applications using these DLLs break when run on Windows 98. While the Version Conflict Manager will save the "newer" versions when Windows 98 is installed, there is no mechanism to tell the user which DLL a program is seeking. Great decisions there!
DLLs sometimes get updated by new applications. But when you test a system, you need to test all possible combinations. When I looked at my system, it had 906 DLLS. If they each get updated only twice that is 3 to the power of 906 possible system combinations to check for stability. The universe won't last long enough to test that. As well as being a solution to software development times, shareable components are also a major problem with software stability.
I changed editors (from SuperNoteTab, after the outline feature switched off) back to The Semware Editor v2.6) but found that if you originally started TSE at a DOS prompt (full screen) that when you started it as a console application in a window, it managed to crash Windows 95.
Then on a restart, ScanDisk found the Windows keyb file was corrupted, and thought a folder had bad long file names. Considering I had never used long file names in the folder, I wasn't impressed. Then ScanDisk crashed Windows. This really is a great system!
Finally managed to get the editor to be well behaved by starting another maximised DOS shell and invoking the editor from that, all via a shortcut, instead of directly. Considering the editor is said to be a 32 bit application, I'm not real sure about that method. As far as I can tell, the Windows console drivers suck.
Ericsson SH888 GSM mobile phone and IrDA ports
The SH888 is one of a handful of mobile phones that includes both an IrDA compatible infra red port, and a built in data adaptor equivalent of a modem. Many mobile phones have an infra red port, but few of these actually run a standard IrDA connection. Connecting a computer modem to a GSM phone makes no sense, since the phone is inherently sending digital signals, and not audio like a regular phone line for which a modem is designed.
I wanted the phone to work with a Psion 5 organiser, for use when I was travelling.
However the phone came with a Windows 95 driver disk, so I thought it might be a good idea to test it first under Windows, before trying to get it working with the Psion. Six hours later, I'd decided it wouldn't work on my Windows 95 desktop system, and I'd had to update from Windows 98 to Windows 98 SE to get it to work with my notebook computer.
In contrast, I just told the Psion I was using a mobile phone, used some sample SH888 settings from the Psion Digest, and two minutes later (and no reboots) I was connected to my ISP, sending and receiving email, viewing the web, and sending faxes.
Tried deleting a subdirectory under Explorer. Got a "path is too deep error". Actually it means the file is cross linked. Mind you, it hadn't been accessed for ages, so I can't see why it was crosslinked.
Eudora Light funnies
Brought home an ftped copy of my email directory from a Unix machine after a trip. Wanted Eudora to read it (just for ease of use), so I renamed it as in.mbx Eudora (and/or Windows) instead decided to zero the file. This happened three times before I finally persuaded Eudora to accept the file. Dunno why it had problems, as Eudora usually accepts "foreign" email files without trouble (unlike Outlook, which uses a proprietary email format instead of simple ascii).
Almost every time I right click and attempt to copy or rename a file using Microsoft Explorer, all my system font starts graying out until I can't read any text on screen.
I made sure I had the latest version of McAfee Virus Scan, but didn't find any external causes. I'm not impressed with having to restart the display every time I want to copy some files in Windows 95 4.00.1111
There is a long rant about Explorer and its problems at www.radsoft.net/resources/rants/ It was once file 20000228.htm
This lists some of the crazy things Explorer does to files, not sorting by extension, getting DOS names wrong by assuming a file name in capitals should just have the first letter capitalised, looking up the registry for everything including icon locations, passing incorrect long file names to programs, mangling file names repeatedly, etc.
In a clean partition, make two directories with long file names, "File Name Long 1" and "File Name Long 2" differing in the last character. The file names will be mangled to FILENA~1 and FILENA~2 Install some programs that need to use the registry in "1". Now install the same (or upgraded) programs in "2". Now delete "File Name Long 1". Now zip up the partition, delete the contents, and restore the contents. "File Name Long 2" is now mangled into FILENA~1, and all the Registry entries that pointed to "FILENA~2" are now wrong! Crock of shit!
First thing you do is get rid of all the idiotic icons. Then replace the Win 98 shell with the Win 95 shell (which isn't much better, but is slightly less bloated). Then stop using long file names, and stick to 8.3 which is less likely to be mangled. Also, stop using any program that needs to use the Registry, if you can (you can't, but every registry entry seems to slow things down).
Turns out that while MS-DOS saves filenames in UPPER CASE, Window NT seems to use lower case, while Windows 95 and Windows 98 seem to use a different method. Mixed case names are displayed This Way, as are names with spaces. Upper case names are displayed with the first letter in Capitals. This is because Explorer considers any name that is all CAPS (that is, any MS-DOS filename) to be an MS-DOS file name. However instead of showing it correctly, Explorer shows it with an initial capital letter, and lower case. In short, Explorer shits all over the file names when it displays them.
Microsoft Front Page seems to generate file names in UPPER CASE by default. Windows doesn't care what case you use anyway, but when you transfer it to a web server, your case must be correct. Most people use lower case on their servers. I think the easiest way to avoid problems is to use an FTP program that converts filenames to lower case when transfering them (the command line one included with Windows 95 does NOT do this - avoid it unless you can change the case of the file!) I wrote a little alias under 4Dos (a DOS program) to convert whole directories of file names to upper or lower case to get around this.
Moshe Frankel of views.com documents a USA Justice Dept case in which he accuses Microsoft of making adjustments to their programs such that they crash Framework and other programs and are sufficiently inefficient that users must upgrade their computers, and that Microsoft ignore and modify standards.
Bill Gates is reputed to have majored in poker at Harvard. You have to be good at misleading opponents, and at bluffing. No co-operation. Drive weak players from the table by grabbing their moeny. Collaboration with standards? Not very likely.
This independent Bulgarian finds bug after bug in Windows, IE and Office. Check his informative site at www.guninski.com/
Grabbed a help file from a CD, a WinHelp32 version of the Hardware Book. This contains helpful details of lots of connectors, cables and adaptors. The sort of stuff that helps when you can't recall where CD is on a 9 pin null modem adaptor. Works great on screen. But when I print a copy, I get garbage. I figured it was my printer driver. Nope, same problem on a different computer and printer. I guess different versions of Windows require different Help File systems. This is no help to me, so I'm off to download the HTML version of the Hardware Book instead.
The free "private edition" of the Hyperterm terminal emulator that is provided with Windows crashes whenever I try to set up a connection under Windows 98 (I finally found an old DOS version of a comms program and used that instead).
My doctor asked me if I knew anything about infra red printer ports, because he had a new Canon BJC80 printer with an infra red port, but his Windows 95 equipped portable computer wouldn't run infra red. I said I'd check on my computer.
A new Micron Transport portable computer, complete with not one but two IrDA infra red ports. Does Windows 95, as installed when you get the machine, include IrDA support? No, it does not. Strike one for the people setting up the computer. Also, there isn't any Windows 95 manual to consult (all on disk, it seems).
So you pull up the Windows 95 help system on your desktop system, which in a Windows 95 4.1111 system doesn't include an entry for IrDA or for "infra" ... Then you run the "find" facility to index the online help, and you still can't get at this information.
However the Windows 95 (original version 4.950) installed on the laptop does have an entry on infra red. It tells you to click on the infra red icon in the task bar. Of course, that icon doesn't exist, because IrDA support wasn't installed.
Better do the testing on the desktop machine, in case something breaks, since the motherboard has IrDA connectors (albeit with no LED or photo transistor, nor any information about what or how to connect them). You can just bring up "Control Panel", and ask Windows to install "new hardware", and manually select infra red.
This would work, if Windows 95 could find the files it needs on the install CD. Since it can't, you manually extract the missing file from the correctly identified place on the Microsoft CD-ROM using PowerZip (since I can never can remember how to use Microsoft's "extract" utility, and "extract" doesn't appear in the Windows "Help" files either).
Having found all that, I installed it on the notebook computer, and once I had guessed which serial port to take over (Com 2), the IrDA port did print to my Hewlett Packard 5MP printer. For reasons unclear to me (considering I'd already installed the Windows 5MP PS printer driver), Windows 95 now wants to install a new printer driver for the 5MP, and wants to do it from the manufacturer's disks. I'm not keen letting it do so, because every other time I've used the HP printer drivers from HP under Windows 95, it has crashed the machine.
Not to make invidious comparisons, but my obsolete Psion 3c palmtop computer, with all of a megabyte of operating systems and applications that do actually do multitasking, also has an IrDA port. It also has a nice little 384 page manual, with 23 pages of indexes and does explain how to use IrDA. Not that you need the manual; the "help" button brought up a list that included complete, detailed and accurate instructions for using the IrDA port, and it worked first time.
Latest report on Windows and IRPort from someone updating their Acer laptop from Win 98 to Win 2000. It seems that Microsoft have stopped supporting IrComm in Windows 2000. This breaks syncing a Psion via IrDA with Windows 2000. Typical Microsoft (I bet they kept it compatible with WinCE).
IrCOMM and Windows 2000
Windows 2000 does not support virtual comm ports nor printer ports using IrCOMM (Windows NT did not support IrDA at all). This means that all your legacy gear, like IrDA printers, mobile phones, and Psion PDAs, fail to work with Windows 2000. Microsoft are telling developers to work with TinyTP via Winsock. I want my IrCOMM support, and refuse to buy an operating system that doesn't support my legacy peripherals. Formerly http://www.microsoft.com/hwdev/infrared/IrCOMM.htm
ISO Language and HTML
Microsoft has done a lot to support languages other than English. However why does it continually break the rules? HTML written in ISO 8858-1 Latin-1 character set have a specified set of escapes for special characters. Microsoft ignore this and insert a variety of non Latin-1 characters between 0x82 and 0x95, thus being incompatible with both Latin-1 and Unicode (extra control characters). John Walker has done some nice stuff about this with his MS HTML Demononiser, to fix things. www.fourmilab.ch/
There is an undocumented Java application viewer "jview.exe" in the Windows 95 directory. No matter what command line options you attempt, it claims it can't find the Java .class file you specify. Borland's JBuilder can run them fine (once you discover that the manual included with JBuilder is mostly wrong about how to run things).
I have a nice small 84 key infra red keyboard, which is real handy for my working methods. Windows 95 is perfectly happy to accept that I have an 84 key keyboard. However, it isn't willing to let me use keyboard shortcuts like Ctrl arrow to move the cursor, nor shift arrow to mark text. Nor does this work on another 84 key keyboard I own. Yet Windows 95 claims it has recognised an 84 key keyboard. Meanwhile The Semware Editor (MS-DOS version) recognises the whole range of keys.
More recently, every keyboard I have has failed after a period of time working under MS-DOS. Just locks up. No clues why this happens. So I installed a mouse driver, and verified it was a general crash, not just the keyboard being locked out. Tried it under Caldera (DrDos/Novell) Open Dos 7. Happens under the Windows 95 MS-DOS 7 but not when that is running as a window or a full screen from within Windows 95. Finally the penny dropped, and I turned off the APM power management control in the board BIOS. That appears to have solved the lockup problem (but the monitor stays up and running all the time now).
Macros in Word
Word uses a .dot template file to save margin, font and type style and other settings relating to a document. If nothing else is there, it uses normal.dot. Ordinary .doc documents don't contain macros, but a virus can rename a .dot file containing a macro as a .doc file. This doesn't happen with Lotus WordPro or Word Perfect files, because the macro file can't be mistaken for the document file.
This wouldn't matter except that Microsoft makes it possible for a macro to run automatically in three different ways. If the macro is AutoExec, it runs when Word starts (you can disable that with a /m option in the command line in your PIF). It can be an auto-macro, of which there are four types. AutoNew, AutoOpen, AutoClose and AutoExit. The virus can attack when any of these acts (you can disable these by holding down the shift key when you open a file). Finally, a macro can be named so it replaces an existing Word command (and you can't disable these).
Best protection. Remove Microsoft Office. If you won't do that, at minimum in MS Office 97 products under Tools, Options, General and tick Macro Virus Protection. At least then you get asked whether to run macros. Also, get an anti-virus program and update it frequently. Office 2000 has better default settings.
The people buying computers today correspond to the people in Cyril Kornbluth's story. Microsoft know it, and rely upon it. There are viruses and worms everywhere, and lots of people to blame for them. However never is the blame pointed at the people who sell the crap that makes this sort of thing possible. Microsoft could make safe and secure products. They have programmers more than talented enough (Windows NT 4 is pretty robust). They choose not to.
The people who buy these products think they are in control, although they don't know enough to make any sort of reasoned decision about programs.
Micron Laptop Battery
Bought a Micron Laptop new in July 1998 from a shopfront dealer. Didn't use the Lithium Ion rechargeable battery very often, but charged it back up every two months or so, discharged it, and then recharged it. In November 1999 I discovered the battery had died. Emailed Micron. Lots of email correspondence, with them saying they couldn't identify the serial number of my Micron in their data base. They gave me several fax numbers for me to send them a copy of the serial number on the notebook. Not one fax number ever worked. I sent them an airmail letter. They said it never arrived. On the other hand, they always did respond to email. In April 2000, they said the battery wasn't covered under warranty (the manual I have says the machine has a three year warranty). They gave me a 1-800 number to phone to buy a replacement. I'm not real sure a 1-800 USA number works from Australia. Sigh. Incidentally, my initial email asked about buying a replacement battery, and the stuff about warranty replacement came from Micron. Now I sort of doubt I'll buy a battery (or anything else) from Micron.
Windows starts and claims vredit.vxd, filesec.vxd and msnp32.dll are missing. Who knows why, when or how they are missing? I edited them out of the registry after a few times because I didn't know what they were used for, and I was sick of Windows complaining.
Windows for Workgroups 3.11 locked up and needed a cold reboot. Turns out a fast 486 or a Pentium is incompatible with a 16550 UART based serial port. Update \windows\system\serial.386 with a new file usually available as wg1001.exe or wg1001.zip. Yes, you use a DOS comms program to get it ...
This is a replacement for IE and Outlook Express, in beta at the end of 2000. The beta adds a 9 line endorsement of how good it is, and will send that to people in your address list. A built in virus! Also, if you log into MSN, Explorer will change your MSN email from POP3 to sending it via Hotmail, and there is no way for you to reverse it! Nor can you prepare your email offline,as the MSN connection will no longer accept Outlook. I'd assume they will change the final version.
I had to install this to rescue a bad drive. Turned out the automatic Windows based front end just made the situation worse. The only utility to help was the DOS based diskedit. On a 4 GB drive, it really could use some better navigation and marking tools. I don't think it has been updated since the days of 20MB drives, and when you have a bigger drive it sucks!
Make a table in Excel, and use a field to calculate the sum of a column. Select Tools, Track Changes, Highlight Changes, Track changes while editing. Change a value in the column. Now see if the total is still correct.
The CALL operator in Excel can access the kernel32.exe and use any API in it, or in any DLL. Internet Explorer and Outlook automatically launch Excel worksheets from web and email, and autorun their CALL statements. Not a macro virus, but it can just as easily destroy your hard drive contents. Netscape does the same thing, but it asks before it runs the spreadsheet, so you have some warning. Just use Windows Explorer, View, Options, select the Files tab, and for EVERY Microsoft file type, every Excel, Word or Powerpoint file, click the "Confirm Open after Download" click box. Then remember not to open any files you download.
Problems are nothing new with computers based on Intel CPUs. Basically the mess started with the design decision to stick with a 64k address space, and use segments to address the actual 1 MB address space on the 8086.
The C language acquired a whole new set of instructions, like near and far pointers, so as to cope with the segmented memory architecture of the Intel 8086 CPU. This idiot idea stuck for nearly two decades.
Other versions of C don't need it, because they run on CPUs that have a linear (flat) address space.
When the 80286 came out, it had two modes, real and protected. The CPU starts in real mode, but if put into protected mode (to address more memory), it has no way to get back to real mode. IBM designers used the 8042 microcontroller (in the keyboard circuit) to keep the CPU reset line asserted for 6uS. Bang, down she goes. They also reserved a byte at address 0x0f in the realtime clock CMOS to indicate the reason for shutdown. See Circuit Cellar #38.
The 80286 enters shutdown mode after encountering multiple errors. (e.g. SP=0001, a Push instruction forces a shutdown rather than wrap SP to ffff and overwrite memory outside the stack). CPU sends specific status output and waits for a reset or a NMI. Original IBM/AT includes a shutdown detection circuit that forces a reset. CPU pops up in real mode, checks the real time clock CMOS for a shutdown code and vectors off. See Circuit Cellar #41 p58.
In the serial port circuit, a receive interrupt on the 16450 (used by RS232 ports between the 8250 and the 16550) may take up to 1uS to flip the summary bit. However a fast CPU can respond to the interrupt, read the interrupt ID (IIR) register, read and process the byte, and check the IIR BEFORE the hardware summary bit changes. But the interrupt bit has changed, so the register is invalid. This forces a modem status interrupt. Circuit Cellar #38
"What the guy is supposed to do is feel uncomfortable and, when he has bugs, suspect the problem is DR-DOS and then go out and buy MS-DOS" Brad Silverberg, former Microsoft VP, in a 1992 e-mail about Microsoft fabricated error messages in Windows 3.1 designed to scare off users of competing DR-DOS. Quoted in "Red Herring Online", August 1998, PC Computing December 1998, p15.
I got the distinct impression that Microsoft didn't feel one of their systems was complete until it prevented running it under any alternate operating system.
Windows 3.1 would run fine under early versions of OS/2. But Windows 3.1 upgrade (ww0981.exe) has several files that are the exact same size as the orginal distribution (GDI.EXE, USER.EXE, SHELL.DLL) but after installing the distribution it won't run under OS/2. No difference in performance noted either. The files that are changed include KRNL386.EXE and COMMDLG.DLL.
Why didn't my Micron notebook PC already have its PC Card slot software installed and working when delivered with Win 95 installed? (The IrDA wasn't working either.) Why doesn't the Card Wizard actually manage to install the appropriate software? Why do you have to install it all manually? When installing it, why does a "Windows" solution go in the config.sys as 16 bit software under DOS? And finally, why after all that does it fail to write to flash or ram cards? The same thing happens with my desktop machine with an Eiger PCMCIA adaptor. (The SRAM cards do work correctly ... under DOS).
A heap of shit. The idea that you open up a set of ports to protect a set of ports just sucks. This is just commercial snake oil. A firewall needs to have a port to the outside world, and a totally separate set of ports to the rest of your network. Separate machine. Not a bunch of vulnerable software on your main machine. Someone will find a way to subvert whichever one is most popular.
Plug and Pray
It seems the pressure to get PCI cards out has lead some designers to ignore device drivers in the plunge to get the product out. Sloppy implementation of the class and sub-class definitions in the PCI configuration space. The BIOS needs this to make the plug and play work. But some vendors just make up the VendorID and DeviceID fields to avoid the registration fees. Thus, plug and pray.
Privacy and IDs
All network interface cards (NIC) capable of connecting to the internet via a TCP/IP network must have a unique 12 digit number, a media access control (MAC) number, provided by the manufacturer. This number is used when routing TCP/IP network packages and is essential to network operation. Thus any networked computer in the entire world can be uniquely identified, although usually only the network administrator cares about these MAC numbers. You can find your MAC number by running \windows\winipcfg.exe.
Microsoft incorporates the MAC number into many of its Windows 98 Registry settings. HKEY_CURRENT_USER\Software\Microsoft\Office\9.0\Common\UserInfo includes it if you install Office 2000 beta. It also appears lots of other places, usually at the end of a line of figures. The number subsequently appears within Word, Excel and Powerpoint documents (Office 97, and 2000 beta), towards the end of the document, after some text _PID_GUID. Each digit is separated by a null (0) character, so they are hard to search for. So each Word document can be tracked back to the computer on which it was generated.
When you register your Windows 98, all the information you type is placed in \windows\reginfo.txt. As well, two headings not present in Windows 95 are included, HWID = and MSID =. These incorporate your MAC number as their last twelve digits. Even if you specify that your hardware information is not to be included, these two numbers are included. When the registration process is complete, the file reginfo.txt is automatically removed (so if you want to check it, do so before the registration wizard finishes). Thus your unique MAC number is included in Microsoft's Registered User Database.
Your MAC number also appears as the last twelve digits of the GUID that appears in cookies from Microsoft, thus showing they are also collected via cookies. The exception would be if you connect via a server, in which case the server MAC number appears.
Mostly all I want to do is to send a file to the printer without any changes. I write a lot of stuff in Adobe Postscript, and another heap in a markup language called QuickScript that uses Postscript for formatting. Windows 95 seems to think its job is to alter my files, when all I want is for it to send them to the printer totally unchanged.
Setting up a "generic text" printer dosen't help, because Windows 95 doesn't actually let output go through unchanged. Luckily there is a free utility called "PrintFile" that bypasses Windows, and that serves well for Postscript files. For QuickScript I don't want to include the 35k of Postscript code that make up the Quickscript routine to every file I edit. Under MS-DOS or Unix I just used a batch file (shell file) that prepends the QS code to my file and sends it to the printer by copying it to the printer port. Turns out that is also the easiest way to handle it in Windows 95 (I put a shortcut to the batch file on the desktop and drag the file to it).
Why do some print jobs from Windows 95 applications never appear at the printer when using the Windows 95 drivers? I never could figure that one out.
Why do some print jobs stall after a few pages?
Why doesn't anything at all print from "list" under "Take Command 32"?
Psion, Symbian and Microsoft
This little bunch come from the Sunday Times, so I wouldn't want to imply you should believe them without checking further. They relate to emails Bill Gates is said to have authored to others in Microsoft, and said to have been revealed by US Government sources during the antitrust trial. www.sunday-times.co.uk/news/pages/tim/2000/05/25/timbizbiz01034.html
Email 28 June 1998 expresses concern about threat to Microsoft from Symbian. Says Microsoft are prepared "to do the most extreme things we can" against Symbian if it worked closely with Sun, a choice he compared with "declaring war on Microsoft". (Psion have no programs that work with Sun systems, to my great regret)
July 1999 email said Microsoft deliberately made some of its software functions incompatible with handheld computers that did not use Microsoft technology (well, we already knew that).
Tried to figure out how to get a Quick Launch bar onto a notebook computer in May 2002. This is the one at the bottom of the display that lets you drag links to it, to keep them off your desktop. I had one on my desktop, same operating system setup, as far as I know (Win 98SE, with Win95 Explorer). Do you think I can find anything of use in getting it to install? The Help doesn't have anything about how to install it. I tried making the same directories (and just why the hell is it in Application data under Internet Explorer?) No use. I checked the registry. The entries are the same. I can't figure out how to get this handy item to exist. (I'm going to set up a bunch of 4Dos aliases so I can use them instead of the desktop - they may not be as convenient, but at least they work!)
If Microsoft think I'm ever going to buy a yet more complicated operating system from them, they should start buying the furnace contracts for Hell!
I think the Windows 95 registry is one of the worst ideas ever foisted on the computing public. Having lots of applications storing their configuration information in a spot ordinary users can neither see nor knowledgably modify, instead of restricting their changes to their own area. One badly written application can trash settings required by others. Having corrupted the registry, it stays corrupted through reboots. Then you keep part of the user profile in the registry, so users have all sorts of fun logging in on other PCs on the same network, and in returning to their original PC. And leave the whole thing undocumented and obscure. Great going! My advice is to strongly favour use of applications that never, never touch the registry.
Log changes to your registry and files using utilities RegMon and FileMon www.sysinternals.com
Some versions of Win95 modify Config.Sys by changing the /p option in the SBIDE.SYS line to the wrong address. Either write protect Config.sys, or get rid of the file that makes the changes namely \windows\system\ctpnpscn.drv Rename it, then edit \windows\system.ini to get rid of all references to it. This seems to be a file provided by Sound Blaster.
SCSI drive won't work
I tried to persuade my PC to use a 1 gigabyte Fujitsu M1606S SCSI drive as a second drive, with an NCR chip based PCI SCSI card.
I tried formatting the drive under Windows 95, and had anomalous results (wrong disk size), so I decided to reformat.
The "fdisk" from both MS-DOS 5 and MS-DOS 6.22 would not run at all, won't open open a display. Caldera's OpenDos Fdisk R1.75 says "error opening system area - can't continue". Power Quest's PQMagic v3.02 does not detect the SCSI drive. Windows 95 claims it is a 2 gigabyte drive (it isn't, it is 1 gigabyte), and then runs into errors around cluster 500 (which is not a lot of formatting).
Installed my older Fujitsu M2684SAU SCSI (528 MB). Same failures in MS-DOS, OpenDos and PQMagic. Windows 95 detects drive at 541 MB (correct) and formatted. Fails at cluster 500 when tested with ScanDisk.
I finally decided the PCI SCSI card is stuffed, so I sorted around my old gear and came up with a Trantor T128 8 bit ISA SCSI card. Does this work? Well, it works fine under MS-DOS with its own drivers, but Windows 95 doesn't have a driver for it, and can't detect my SCSI drive.
Finally got back my Adaptec 1520 SCSI controller. This includes its own drivers in ROM. Works under DOS. Works under Windows (Windows claims there is an interrupt conflict - there isn't, as the SCSI card is set manually by jumpers, so I had to tell Windows that I don't have a SCSI card if I want the SCSI card to work ... go figure!) But, it won't work if it is above drive F: and since I have multiple partitions on my IDE drive, I can't run my secondary IDE drive in the same boot session as my SCSI. One of them has to be removed!
Socket 7 vs Slot 1
From about 1993, Socket 7 motherboards were built by myriad companies. As a result of this competition, prices for motherboards tended to be low. Intel (largest CPU maker) have tried to move the motherboard on to a proprietary design, with their Slot 1 ( and Slot 2, and whatever the slot for notebook computers is called). Companies using these designs have to pay Intel for the right to use the Slot 1, and Intel initially were the only company making support chips. With new Socket 7 designs accessing the board at 100 MHz, for a single CPU board (which is what most end users buy), there is no great advantage to a Slot 1 board. And the price of a Socket 7 board (and a non-Intel CPU) is considerably lower.
Unless you have real good reasons for paying more, and actually want to be locked into a single vendor solution, I'd suggest your motto should be "Intel OUTSIDE". If Intel put Socket 7 compatible CPUs on the market again (they withdrew them in 1998), I see no reason not to use them - Intel make reliable CPUs.
Updated May 2000. Seems even the Super 7 socket is now dead, with AMD no longer making CPUs to fit. I'm not planning to update my 1997 computer, so I haven't tried to follow current model motherboards.
After solving the mystery of my desktop machine crashing under MS-DOS but not under Windows, I came across a Mobile Computing mention of the standby, suspend and resume, and hibernate features in notebook computers. In November 1997 they tested 10 notebooks, and found an average time to cold boot of 51 seconds (best time 35 seconds, worst time 71 seconds). Resume from disk averaged 32 seconds, resume from memory averaged 15 seconds. However two froze on resume from memory, and one on resume from disk.
For reasons unknown, many of the programs listed under Programs and in folders off Programs just disappeared one day. I have no idea why or how. I found this when I went looking for Hyperterminal to test a modem. Gave up and installed Kermit and tested the modem under DOS instead.
Toshiba notebook service
Journalist David Forman, stranded in Silicon Valley with a month old crashing Toshiba notebook, finds three technicians who blame the problem on Windows 95's self destructive tendencies. International warranty doesn't help, because Toshiba claim the problem is Windows. However Toshiba were the ones who supplied Windows pre-installed! If it doesn't work, why supply it?
Tried to use add programs to uninstall Marketscape's WebCD viewer. Uninstall silently fails to remove anything. Sausage Software's Hot Dog uninstall simply claims it can't do it.
The Office 97 service release 2 came out in September 1998, and is a 23 Mbyte file. You know what the chances of downloading that from Microsoft without a failure is? And you can't get just get a Word update, you have to get the whole thing. There is a free CD version, but the one they send you by default absolutely requires that first you install SR1. If you insist (it is not mentioned on the Microsoft site) you can get the "full install SR-2 CD option" (called SR-2 Full CD in Australia) which doesn't require any patches, but you do have to uninstall Office and reinstall. If you buy a new copy of Office, you get the old version, without SR2 patches. If you reinstall any subset of Office, you then have to reinstall SR1 patches before the SR2 patches can work. You have got to be kidding!
Opps, Microsoft appear to have withdrawn the SR-2 patch due to problems with its install (they eventually fixed that).
USB - "Polishing a turd"
The "polishing a turd" was a comment by Maury Wright, technical editor with EDN in the July 8 1999 edition, page 44.
Intel pushed USB, claimed it would get version 1 out in 18 months. It took 4 years, and still doesn't always work. Intel now say they need version 2, and they want it a lot faster than the nominal 12 Mbps.
The computing world should be heading towards distributed intelligence, and compatible communication. A camera and a computer should be peer to peer, for example (or do you think it reasonable that a camera be too dumb to send a photo to a printer?) Which is the host when your smart PDA connects to your smart PC? USB is entirely PC centric. USB demands the host initiate all transactions, and as soon as you put multiple devices on it, things slow way down. I don't believe it will be suitable for fast mass storage devices.
Better if they instead moved to IEEE 1394 (Firewire), which is already faster, and designed for peer to peer operation.
The Norton anti virus folks say in the two years since 1996, virus infections became 20 times more likely. Well, they would say that, wouldn't they. No longer confined to executable viruses and boot sector viruses, now you get Office macro viruses (actually, you can get macro viruses for anything with a sufficiently complex macro language and an autorun "feature"). You can also look forward to Active X and Java viruses (my attitude is to turn both off).
Simple solutions to the virus problems. Never boot from a floppy you obtained anywhere else. Never run a program someone sends you. Turn off all scripting. Get rid of MS Office, IE and Outlook. No virus problems.
Richard Smith has been a gadfly all through his days at Phar Lap Software, with repeated reports of Windows bugs and privacy violations. He reported on the cute trick of Office putting your Ethernet card MAC number into Office documents, thus allowing your Office material to be traced to your PC. Towards the end of 2000, he showed how these tricks can be used to track all the recipients of an Office document. He has moved to the University of Denver Privacy Foundation.
Web bugs are small almost invisible pictures on web pages that link to other web pages. Quicken and FedEx use them. If say a Word document includes a link to a picture, it can go on the web and download it when needed. So your browser sends a return address (as it must to receive the picture). Then it says what sort of browser it is, version, and the operating system. Then it sends a "referrer", the location of the last web page you browsed, then some misc stuff. So if you want to find wo is reading a document, embed a link to some small picture (different one for each document), and when a browser comes after that picture, capture all the details of that download (these are in your server log files anyway). Works on any web enabled document, any html email. The picture can be a single pixel in size, almost invisible. Now the address that you supply will be a permanent one if you have DSL or cable connections, but if you go via a modem to an ISP, it will be one of several hundred assigned dynamically (and may be different tomorrow - but can be tracked via ISP logs). (This is not a Windows problem).
If this page crashes your Windows 95/98 system, check Microsoft Security Bulletin MS00-017 for a patch for "DOS Device in Path Name" at http://www.microsoft.com/technet/security/bulletin/fq00-017.asp (I've removed this, as it is obsolete.)
In the July 1997 Scientific American there was an article Taking Computers to Task. In part, it compared the complexity of various versions of Microsoft Word, by equating area with the number of commands. Word 2.0c has 311 commands, while Word 97 has 1033 commands. Ron Kmetovicz commented in Electronic Design September 15, 1997 that he thought and typed at the same rate using the word processor in Works for DOS. He experimented with his productivity over a period of time. "The new versions are no better, or worse, for creating original text." "Simply put, for me, the new tools do not improve my productivity when creating original material, and they really offer minimal ease-of-use benefits". He concluded the only reason he had newer versions was people kept sending him word processor files that he couldn't use, because they had moved to newer versions.
The funny thing about Works is that you can't move to Office. Word doesn't recognise Works .wps files (needs a conversion file). Excel and Access don't recognise Works .wdb files, and there is no conversion file. Microsoft say use comma delimited, but this doesn't work because large text files from Works are cut off.
Comments about Microsoft
"Microsoft has generated a machine culture where every user is on their own; where every single machine is set up differently, making support a nightmare; where there is no adequate support for software; where work is lost; where systems crash; where programs that work fine today will fail tomorrow because of some apparently unrelated change to the system; where programs are created to be just adequate; and where simple tasks are simple, but complex tasks are accomplished by people doing simple tasks again and again because there is little or no automation available to users."
"And the people purchasing systems have bought it, because it all looks so easy. They were sold the idea that you could get some cheap software that would apparently do some simple jobs, but they were not told it would cost them significant sums in human times to get around the problems that the software would induce." Peter Collinson, Sun Expert, Feb 1998
Where Microsoft Gets Its Programs
Microsoft often buys programs or parts of programs from other companies.
- Access database
- FoxPro database was bought in 1992.
- details unknown
- Access Software Links Gone
- Softimage 3Danimation bought in 1994 (sold to AVI in 1998).
- details unknown
- Aspect Engineering
- details unknown
- Blue Ribbon 1995
- Business Solutions
- Axapta Accounting purchased by Naviision, then Navision purchased by Microsoft. Soloman purchased by Great Plains, then mid market Great Plains accounting purchased by Microsoft and now part of their CRM.
- Cabinet (.cab) files
- Jonathan Forbes gifted them with his LZX compressor, with which they invented a proprietary packaging system, called Microsoft Cabinets, or .CAB format.
- Purchased nCompass cms in May 2001.
- LinkAge Sotware Canada in 1997 for connectivity software.
- file compression
- Stac sued and received US$120M
- details unknown
- Flight Simulator
- Bought from Bruce Artwick organisation.
- Microsoft bought HTML and page editing company Vermeer Inc., of Cambridge MA. for US$132M in January 1996.
- Microsoft bought Bingie in 2000. Bioware role playing games in April 2004. Digital Anvil company purchased in 2000.
- details unknown
- Purchased outright.
- Internet Explorer
- Licence from Spyglass.
- Bought rights to use QDOS (quick and dirty operating system) from Seattle Computer Company.
- Visio purchased in 1999, made part of Microsoft Office.
- Developed by Bob Gaskins in 1984. He joined Forethought, and did Presenter, which changed name to PowerPoint. Powerpoint 1 came out for Macintosh in 1987. Microsoft bought it in 1987 for $14M. Came out in 1988 for DOS and planned for Windows.
- Written by Richard Brodie for IBM PC DOS in 1983. Charles Simonyi of Xerox Parc joined Microsoft in 1981 after creating the first GUI word processor Bravo.
- Subject of a patent infringement case by Wang.
- Lookout add-in for Outlook bought in July 2004 for use in MSN Search.
- Pocket PC
- Bought Intrisync of Vancouver BC wireless PDA microPDA reference design in December 2002.
- programming tools
- Colusa object oriented programming bought March 1996.
- SQL Server
- Based on Sybase code.
- Interix Unix subsystem on Windows for migration purposes.
- Virtual PC
- Runs Windows on a Macintosh, bought Connectix, who made the product. Now part of the Pro version of MS Office for Macintosh.
- Visual Basic
- Code purchased from Cooper Software.
- Visual C++
- Purchased Lattice C compiler.
- Some licences from Apple, multitasking from Dynamical Systems.