os2fan2

There are some things to note about UMBs on modern machines and Windows. 1. You should avoid using UMBs under Windows, since this slows it down somewhat. Have a look at UMBPCI on the web on this. 2. Modern machines use more UMBs then the average 486, so there's less for DOS. UMBPCI will find out what's available. I generally try to avoid WFW311, sticking with Win311. Still, i don't think there's a lot of changes.

Some of their utilities are neat. They have ctm, a console-mode task mamager, and the freedos command.com (cmd.exe) and edit.com (edit.exe) recompiled to Winnt. You can get things like the device manager and user manager to function without all that other MS legacy attached to it.

I had a look at the IBM thinkpad disks (which are basically recovery disks, but the process can be done from PE). The base copy of Windows is an unmodified source tape. The modifications come between the text-mode install and the gui. You install windows in some sort of other mode from PE., using the switches in winnt32 to point to the correct boot drive. Once the source is in place, you copy over files as needed (usually oembois.*), and then on the reboot, the install happens normally. Usually the OEM toys are added after the setup, which is why you don't find things like OEM bitmaps in Windows or DOS setups.

Compress v 1.02 and 1.11 will create KWAJ files. These date from 1990 (the WLO10.ZIP file for OS/2 is compressed in these). It's used in DOS 6, and Windows 3.11, it's hard to find. Vers 1.11 is the go, because the default rename is to edlin.ex_ rather than edlin.ex$. I got 1.02 from the fox pro 2.6 redistribution, and 1.11 in the msdos 6 source code package. Compress v2.0 is available in the Windows 3.1 SDK and many other things, that produce SZDD files. The compress in the Windows NT 3.10 reskit produces this file. Compress v2.5 in the resource kits, etc, by default, produce MSCF files (MSZIP), with varying degrees of compression. Look for CABPACK v 1.1 for a DOS version of this. Expand 2.1 (windows 3.11), expands both KWAJ and SZDD, but not MSCF. Other expand utilities work with some subset of SZDD, KWAJ and MSCF. The DIAMOND utility even in the earliest packages, are Win32. You could look for cabpack 1.1 for a DOS version. Needs to have the Runtime 200 fix applied, though.

Here are some things ye might like to try. CONFIG.SYS files=50 buffers=30 STACKS=9,256 device=c:\dos\himemx.exe /max=64512 /x REM device=c:\dos\emm386.exe noems i=b000-b7ff /verbose dos=high,umb,NOauto device=c:\dos\cdrom.sys /d:satadvd lastdrive=z You don't really need AUTO here, because you're loading stuff through config.sys Try UMBPCI, http://www.uwe-sieber.de/umbpci_e.html it gives UMB's (ie dos=umb), without loading EMM. AUTOEXEC.BAT @echo off path c:\dos;c:\windows set temp=c:\temp SET TMP=c:\temp PROMPT $p$g lh mscdex /d:satadvd /l:x REM lh ctmouse REM idle If you are just testing whether windows will load, try removing the mouse and idle. Add the line 'pageovercommit=1' to the 386Enh section of System.ini in your windows directory Try using vga driver or vga30 driver, to see if this cures problems. SYSTEM.INI [386Enh] ... PageOverCommit=1

You might be able to copy the files through a share from a PC with a working cdrom. You could use cross-over network cables, or even something like 'interlink' cables, which are cross-over patch cables. The alternate is to extract the hard disk, and load the files onto it through an external hard drive. This is how i usually install OS/2 2.x on vm machines (where the cdrom is not seen by the install).

BennyLevel is a hack around the DOS errorlevel bug, which treats a two letter errorlevel as 10*h- + l - 11*'a'+11, modulo 256 (c2d('x')*10+c2d('a')-528)//256 (c2d('H')*10+c2d('a')-528)//256 (10*('x'-'0')-('$'-'0'))mod 256 lower case (10*('H'-'0')-('$'-'0'))mod 256 upper case What the batch does here, is test the hex on HA @ECHO OFF choice /n /cABCDEFGHIJKLMNOPQRSTUVWXYZ Choose drive: : FINDCD FOR %%D IN (A B C D E F G H I J K L M N O P Q R S T U V W X Y Z) IF ERRORLEVEL H%%D SET DRIVE=%%D FOR %%D IN (a b c d e f g h i j k l m n o p q r s t u v w x y z) IF ERRORLEVEL x%%D SET DRIVE=%%D ECHO You chose drive %DRIVE% ================================================ Benny Pederson recently discovered an important new batch feature (4th November 2000 in alt.msdos.batch thread "Choice" ), namely that ERRORLEVEL checks can be made with characters other than 0-9. As I posted at the time, the values of the characters run in parallel to their ASCII values (they are their ASCII values less 48 decimal, 30hex). In other words, Command.com _doesn't_ check range on the character, it just subtracts the offset ASCII character 0 (zero = 48 decimal, 30 hex). To distinguish this from ordinary ERRORLEVEL checking, I call this _BennyLevel_ checking. For convenience, throughout the following account, I call ASCII-Value-Of [character] - 48 the BennyValue of the character. All normal keyboard characters can be used (except the five in exception list below): from the character next in ASCII order after [space], namely ! (exclamation 33 decimal, 21hex) through to ~ (tilde 126 decimal, 7Ehex). High ASCII characters (above 127 decimal, 7Fhex) can be used also, but because of code page variations, this is less useful. In the rest of this, I use DECIMAL arithmetic to avoid the constant translations in parentheses. A very small number of characters can't be used because of conflicts. The five characters which can't be used are: comma , semicolon ; less than < equals = greater than > Note that percent (%) _can_ be used (but _beware_ of generating internal script conflicts with environment variables) For combinations of two Benny characters, Command.com just applies the algorithm: if second character present, multiply BennyValue of first by 10, add BennyValue of second. Similarly for three characters: (BennyValue first x 100) + (BennyValue second x 10) + (BennyValue third). To give examples: The BennyValue for A (ASCII 65)= 65 - 48 = 17 decimal, thus IF ERRORLEVEL A is true for ERRORLEVELs >= 17 decimal The BennyValue for a (ASCII 97) is 97 - 48 = 49 decimal, thus IF ERRORLEVEL a is true for ERRORLEVELs >= 49 decimal Also _note_ that ERRORLEVELs work modulo 256, so IF ERRORLEVEL 513 and IF ERRORLEVEL 257 and IF ERRORLEVEL 1 are all equivalent. BennyValues make it very easy to construct high ERRORLEVELs. In my opinion, this is a _very_significant_ and exploitable feature. It (turns out that it) has been a feature of Command.com for the past 14 years (since DOS 3.20) for certain, and almost certainly since DOS 2.0 when the IF command and ERRORLEVEL checking were both introduced. As far as I know Benny Pederson was the first to discover this feature in all this time. Simple example showing calculation to simplify Choice menus: The first of the two example batch scripts I posted was: ::====BENNYDRU.BAT @ECHO OFF choice /n /cABCDEFGHIJKLMNOPQRSTUVWXYZ Choose drive: FOR %%D IN (A B C D E F G H I J K L M N O P Q R S T U V W X Y Z) DO IF ERRORLEVEL H%%D SET DRIVE=%%D ECHO You chose drive %DRIVE% ::==== Only four lines in the above (watch for accidental line wrap). In this script, BennyLevel checking is used with an offset (calculated as below) to make the Choice letter result and the corresponding BennyLevel check letter from the FOR IN DO loop become the same letter in each case. To make this work, you need the ERRORLEVEL resulting from each Choice entry to match the BennyLevel check by that letter. Choice is not case sensitive, but BennyValue _IS_, of course, case sensitive, because it derives from the ASCII value. In this example I work with an Upper Case BennyLevel check The BennyValue of A is 17 (ASCII 65 - 48). A (or a) in Choice in the above returns ERRORLEVEL 1. To make them match, we use ERRORLEVEL overflow - remember ERRORLEVEL works modulo 256. Also remember the algorithm Command.com uses for character pairs is 10 x BennyValue(first) + BennyValue(second) So we match ERRORLEVELs 17 and 1 by solving the following equation in integers m and n: 10m + 17 = 256n + 1 and it's easily seen that m = 24 and n = 1 will solve this. The important part of the solution is the m value. The character whose BennyValue is 24 = H . Therefore, the BennyLevel check: IF ERRORLEVEL HA is the same as IF ERRORLEVEL 1 and similarly HB is the same as 2 and so on. Hence, the offset H in the above Uppercase Benny check. Similarly, we can produce a lower case result as in the second script: ::====BENNYDRL.BAT @ECHO OFF choice /n /cABCDEFGHIJKLMNOPQRSTUVWXYZ Choose drive: FOR %%D IN (a b c d e f g h i j k l m n o p q r s t u v w x y z) DO IF ERRORLEVEL x%%D SET DRIVE=%%D ECHO You chose drive %DRIVE% ::==== Only four lines in the above (watch for accidental line wrap). Here the BennyValue of a is 49 (ASCII 97 - 48), so the equation to solve becomes: 10m + 49 = 256n + 1 We easily see that n = 3 and thus m = 72 solve this. If you can't see this quickly, all I am doing is finding the first multiple of 256 that when you add 1 gives a last digit of 9. Then I know that when I subtract 49 I will get a multiple of 10 (in this case 720). Again, the important value is m = 72. We need the character whose BennyValue is 72 = x (ASCII 120 - 48). Hence the offset x will allow a simple FOR IN DO lowercase Bennycheck of the ERRORLEVEL from Choice without any padding characters. All the padding can be done with a BennyValue offset prefix to the ERRORLEVEL check. You would use the first version of the script (with Upper Case list in FOR IN DO loop and offset H) if you wanted to return an Upper Case result. You would use the second version of the script (with Lower Case list in FOR IN DO loop and offset x) if you wanted to return a Lower Case result. The case of the Choice list doesn't matter in either version, of course, since Choice is not case sensitive. Further note on characters prior to ASCII 0 ( = 48): =================================================== Characters prior to ASCII character 0 (=ASCII 48) work as follows: Take example of ! (exclamation = ASCII 33 decimal). The BennyValue of ! is 33 - 48, but Command.com works modulo 256 so this is translated to 33 - 48 + 256 = 241 Thus: IF ERRORLEVEL ! is equivalent to IF ERRORLEVEL 241 and, because of the properties of modulo arithmetic, these characters work in combinations with other characters, too: thus, !A is the BennyLevel for 123, so IF ERRORLEVEL !A and IF ERRORLEVEL 123 are equivalent. Work this out as follows: BennyValue of ! (ASCII 33) = (33 - 48) -15 = 241 (mod 256), (to get the value mod 256, just add 256 to -15) and BennyValue A (ASCII 65) = 65 - 48 = 17: hence: 10 x 241 +17 = 2427 = 123 (mod) 256 ====== And all this has been in Command.com since DOS 3.20 and probably since DOS 2.0. All thanks to the original programmer nearly 20 years ago who _didn't_ put in an entirely unnecessary range check<G>. -- William Allen This works under command.com, but not under 4dos.com.

DOS findings by os2fan2 Here are some interesting things i found and collected of dos. Drivers Emm Ram Smart Pk Himem 386 Drive Drv Mouse Mscdex MSD Win286 2.11 C 880527 1.0 2.04 1.05 Win386 2.10 C 880907 2.04 2.10 2.10 DR-DOS 3.40 C 890630 ? R1.01 MS-DOS 4.01 C 890407 2.04 4.00 2.12 2.10 DRDOS 5.00 900814 OS/2 2.11 940129 2.10 4.00 (9.01) MS-Win 3.00 2.60 4.10 3.04 3.03 7.04 B PDS 7.10 U 900624 2.60 3.04 3.03 MS-Win 3.01 S 901031 2.60 4.10 3.04 3.06 7.04 MS-WMM 3.07 C 911001 2.60 4.10 3.04 3.06 1.10 IBMDOS 5.00 S 910509 2.77 4.20 3.06 3.13 7.04 OS/2 NT (2.77) (8.00) (2.21) 2.00 MS-DOS 5.00 S 911111 2.78 4.33 3.06 3.13 IBMDOS 5.02 S 920901 2.78 4.33 3.06 3.13 8.20 2.00 MSC 7.00 K 920305 2.78 4.33 3.06 4.00 8.20 2.00 b 068 3.10 920203 3.03 4.41 3.06 4.00 8.20 2.00 x MASM 6.11 K 920305 3.07 4.44 3.06 4.00 8.20 MS-Win 3.10 S 920310 3.07 4.44 3.06 4.00 8.20 2.00 MS-WfW 3.30 S 921001 3.07 4.44 3.06 4.00 8.20 2.21 2.00a DR-DOS 6.00 C 920407 MS-Win 3.11 K 931231 3.07 4.44 3.06 4.00 2.00 MS-DOS 6.00 K 930310 3.07 4.45 3.06 4.10 8.20 2.22 2.01 IBMDOS 6.00 K 930629 3.09 4.45 3.06 4.10 8.20 PC-DOS 6.10 K 931116 3.09 4.45 3.06 4.10 8.20 MS-Fix 6.0a Z 930601 4.20 PC-DOS 6.30 K 931231 3.09 4.48 3.06 5.00 9.01 2.23 MS-DOS 6.07 C 940101 3.10 5.00 3.07 5.00 2.22 2.01 MS-DOS 6.20 K 930927 3.10 4.48 3.07 5.00 8.20 2.23 2.01 MS-DOS 6.21 K 940213 3.10 4.48 3.07 5.00 2.23 2.01 MS-WfW 3.31 K 931101 3.10 4.48 3.07 5.00 2.23 2.10 MS-DOS 6.22 K 940531 3.10 4.49 3.07 5.01 8.20 2.23 2.11 DR-DOS 7.00 C 940126 PC-DOS 7.00 P 941117 3.15 4.50 3.10 5.10 8.20 2.25 DR-DOS 7.03 C 980107 PC-DOS 7.10 C 031002 3.15 3.10 5.10 2.25 Win-95 3.95 X 950711 3.95 4.95 3.06 5.00 (8.30) 2.25 2.13 Win-98 3.98 X 990423 3.95 4.95 3.06 5.02 (8.30) 2.25 2.14 Win-ME 3.99 (3.99) 4.95 3.06 5.02 (8.30) 2.25 2.14 Windows 3.01 and 3.07 are 3.00a and 3.00 mme respectively. Wfw 3.30 and 3.31 are vers 3.10 and 3.11. MS-DOS 6.07 is MS-DOS 7.0 beta 1. It is largely based on MS-DOS 6.00, but has drivers of version 6.20 vintage. None of these drivers seem to be dos dependent, save for MS-fix for smartdrv,exe 4.20, which had an artificial check for DOS 6.0 or greater. It actually requires DOS 3.3 A number of utilities that ship with DOS can be used with other versions of dos. 1 deltree, move and choice do not check for dos versions, and work with DOS 5 2, acalc, backup, crc, dynaload, internk/intersvr, msd, qconfig, xdf, xdfcopy, work with any version of DOS. Many of these come from PC-DOS 7 (which largely have the versioning removed) 3. DOS-free versions packages that work with DOS. defrag and scandisk complain for versions of dos less than 6.0 m=msdos, p=pcdos. m63=msdos 6.21, m64=ms-dos 6.22, p52 = pcdos 5.02 ADOS m60 m62 m63 m64 CPBACKUP p60 p61 p63 p70 CPCOMMON p60 p61 p63 p70 DBLSPACE m60 m61 m62 DOSSHELL m50 p50 p52 m60 m61 m62 m63 m64 p60 p61 p63 p70 DRVSPACE m64 m70 m71 E3-DOS p60 p61 p63 p70 EDITv2 m70 m71 FILEUP p70 IBMAV p60 p61 p63 p70 MEMMAKER m60 m61 m62 m63 m64 m70 MSAV m60 m61 m62 m63 m64 MSBACKUP m60 m61 m62 m63 m64 m70 NETWORK m60 m62 m63 m64 PCMCIA p60 p61 p63 p70 PENDOS p60 p61 p63 p70 QBASIC 1.0 m50 p50 p52 QBASIC 1.1 m60 m61 m62 m63 m64 m70 m71 REXX p70 SSTOR p61 p63 STACKER p70 VIEW p70

I'm currently working on a rxloop script for the post rollup fixpack. This consists of a csv-style table, which is run to generate the output. A possibility is to read the [Files] section in dosnet.inf etc to get the assorted working lists. Because rxloop allows you to build scripts, you can keep the hotfixes in their original order, and select the files directly from unpacked hotfixes. At the moment, i am halfway through the rxloop table. There are several different setup routines, some of which have registry patches. These are being saved as well. I am not sure whether to write these as registry files or prepare a rexx registry processor to do this. The rexx registry has the advantage of being able to read registry, and also write reg_expand_sz strings without having to do unicode binary stuff. The data file could look like this. cdd also will make the folder if it does not exist. it depends on w32utils.dll behaviour. cdd hklm\system\.... dword name=value (decimal or hex) string name=value none name delete name expand name=string

In windows 3 and 4, the registry is stored in a prototype, which can be loaded into registry, So there is a file in there that becomes the default, system, software keys. This is already known, the plan is to write a rexx command processor to handle this, something like cdd [key] dword name=value expand name=value etc

If it's a cpl icon, you can make a shortcut to it, and it really does open the dialog box for you.

I've revived the old tapes saved to cdrom. I've been off line some what, and the system had a few crashes. still it was all on a diskette some time back, so that's good, too. MODS: most searches point to this thread, so it's best to allow me to revive it - rather than locking it out because it's old. Please don't lock it! The current intent is to build a post-rollup, for which one expands the sp6a and the rollup, and then run assorted supplied scripts to integrate all of the stuff into the winnt source files. There's a couple of extra fixen existing, which are part of DOS (like the phatcode qbasic fix), which i plan to sneak in as well. They're mostly fixes, not new functionality, although some consideration to be given to EDIT along with QBED (edit.com v 1.0 or 1.1). Maybe even a DOS help file for Windows nt4! And, yes, it really does work without having to apply sp6a or anything.

Are you suggesting heading back to the Windows 2.x 3.0 days, where the control panel was unified? A lot of the control panels can be accessed from elsewhere. None the same, you will eventually find what you are looking for in the control panel. I can't see the advantages of integrating 'file-tools-options' of every installed application into the control panel. Some programs do have a place there. In my control panel, there are some 'hidden' cpls, some driver and application installed applications, and a number of home-made options (the three registry programs). A number of cpls are downloadable from msft, (desktop themes, directx, text services, and tweakui). None the same, the control panel is still stuck in the windows 3.1 progman days. Worse, actually. You can't create a folder in the control panel, and move icons there. Even if you do manage to get things into admin tools, if they're .cpls, they still show in the control panel. You can't for example, create a folder 'input options', or 'woofies' or 'driver icons'. Some installs add icons to the panel, while others add extra tabs to the icons, or replace the icon concerned. I suppose this was the intent of GUID's, but some ridiculous things have GUIDs. Still 'folder options' on a guid is pretty awful, when it would be better served by folder.cpl or something. The MMC (microsoft management console) in my view, is a fairly silly idea. OS/2 went that way (where FDISKPM was replaced by a java app!). The idea is that you need to load some large app to do things like manage users and disks, the whole set is not user managable. User management is now scattered over three or four different applications, none of which gives complete service. It's better to have one icon.