wierd_w

Some potentially useful info to stuff into your static bag with it: https://stason.org/TULARC/pc/hard-disk-floppy-controllers/C-D/DATA-TECHNOLOGY-CORPORATION-Two-IDE-AT-or-EIDE-int-288.html

Looks like a standard part for the time to me. PIO mode 4, 32bit transfer mode. Like other cards of its era, predates bus mastering, so no UDMA modes. It is otherwise an unremarkable IO controller. Does have a miniport driver for windows though: http://vogonsdrivers.com/getfile.php?fileid=1683&menustate=0

Oh, ANOTHER thing you can do, if you KNOW FOR SURE that a pernicious bit of garbage option rom is present (and you would like to get rid of it), is abuse how EMM386 does the UMB mapping. Specifically, EMM386 uses the MMU from protected mode to literally UNMAP whatever *WAS* there, and then MAP IN some RAM from XMS. This means we can use it to obliterate that option rom-- like so: There is a caveat to this though: You *CANNOT* put the pageframe where the option rom was. EMM386 will complain about an option rom being in that location. EMM386 apparently expects a region that is not mapped to ANYTHING to use as the pageframe location. Additionally, if there are several option roms, or an option rom and a segment of unmapped memory, etc-- that are right next to each other, you cannot just define the whole range with an include statement. Each individual block of "something" needs its own include. Like pictured. (I=E000-E1FF is for the bit of ESCD RAM we are using as UMBs, and then I=E200-EFFF is where we obliterate that option ROM.) If you try to just lump them all togehter, eg-- I=E000-EFFF, EMM386 will error, and tell you "Ranges overlap". Forcibly assigning RAM there to use as UMBS? Totally allowed! No complains! BUT YOU MUST ABSOLUTELY BE SURE that you DO NOT NEED that option rom. I have configured the system to use MSCDEX again, (since with that option rom out of the way, we have plenty of room for it), and loaded smartdrv in upper memory. That is why all the UMB is filled. This is useful if you have a pernicious PXE bios in there that you will NEVER EVER USE, Dos NEVER EVER calls into it, and for whatever reason you cannot disable it in the BIOS. You can abuse how EMM386 does the mapping to forcibly map it out.

You know what. It wasn't requested, but I will give a mini how-to here anyway. Quick and dirty primer on manually massaging DOS memory, with MSD. STEP 1: Boot DOS without any memory managers loaded. We want to see the naked, unmolested memory map. As you can see below, this is JUST DOS with nothing else loaded or running. We want to see what the memory map looks like. MSD will tell us. While this is inside a virtualPC virtual machine, it DOES give us a more or less accurate depiction of what a real memory map would look like. As you can see, we have an option rom in the adapter region between C000 and C7FF. This is a very sizable chunk (~32k), and this is no bueno. This is likely caused by the virtual machine emulating a "newer-er" pentium III chipset. If there was a way to make virtual box emulate a VX chipset early pentium, it is what I would be using here, as the memory map is MUCH cleaner. Incidentally, this ugly map is more real to life, so let's just roll with it. This is why you should be more attentive to what chipset you are using, and have some inclination what to expect with different ones, when deciding on a retro computer build. I am a big fan of early pentium VX boards, as they are "Sufficiently fast" and also "Sufficiently slow", that you can get the majority of games to work with little issue, while having a very open adapter region. (Upon further inspection, this 32k option ROM is in fact the video bios.) Additionally, we can see that MSD sees "maybe free" between E000 and E1FF. This is most likely ESCD data, which is technically RAM that contains information about plug and play cards, etc-- and is a feature of such "newer-er" boards. It may or may not be possible to reclaim this area for UMBs, but if you do, DO NOT load windows. In strict parlance, this area is owned by the system BIOS, and should not be walked on. Windows tries to read this area (ESCD) to configure its plug and play implementation, (since the bios advertises ESCD), and if you have programs in there, instead of ESCD data, well, you probably will get what you deserve for trying to be clever. There is additionally an option ROM in memory at E200-EFFF. I have no idea what this option ROM does, but it makes life more difficult by being there. For DOS only, it might be OK to try and reclaim this (ESCD area)-- because DOS does not use this location for anything at all. This memory map tells us we have an unused gap between C800 and DFFF we can use for UMBs. If we set up an EMS page frame, this is not much to work with-- maybe 32k of UMB? That out of the way-- let's do this the "no frills" way, without any special tricks. Just a CDrom driver, a mouse driver, and some memory managers. With only 32k of UMB, the CDROM driver alone probably wont fit up there. And as you can see, NO, the OakCDrom driver does NOT fit into the remainder of the UMB free, after DOS loads itself high. It is FORCED to load into conventional memory. After some trial and error, if we enable the ESCD area for UMBs, we can get DOS and EMM386's page frame to play nice with the ATAPI driver, by putting the pageframe at C800, and including the ESCD area. So far so good. Let's see about getting MSCDEX loaded... And, as we can see, there is no way in hell the MSCDEX driver is going to in the tiny 6k of UMB left... Let's try SHSUCDX instead. Well look at that! It FITS! (barely.) CTMOUSE uses a whopping 3k, so letting it live in conventional should be fine. We dont have 3k free, just some random bytes free in UMB. We could STILL get it into UMB, if we include the monochrome video memory area at B000-B7FF, but then monochrome games will fail spectacularly. Sacrificing the 3k conventional seems like the best compromise. With that dangerous option turned off, we have 615k free. If we turn it on, we should have quite a bit more free UMB area, and we can attempt to load the command interpreter high using some actual voodoo (highlighted)....

Unless you explicitly need it for something, it is very hard to justify the extra weight of drivespace(3). It horks down 64k of ram all by itself. Removing the .bin files from the root is something i tend to do straight away, just for that reason. Mscdex is also a hog. Eats 10k all by itself. Shsucdx is a bit fickle, but uses a lot less. http://adoxa.altervista.org/shsucdx/

While not vlb, an etherlink iii isa card might be fun. Its a 10mbit ethernet card, with good dos support. Sadly, dos packet drivers eat a lot of memory. However, having ipx/spx for retro lan play could be a nice option.

A scsi card would live in the adapter rom region. This would adversely affect the amount of UMBs that can be created, and thus, the free conventional after bootup. ASPI drivers would also need to be loaded. A VLB ide controller might be doable though.

You can control ctmouse's loading behavior, if you know exactly what kind of mouse, and where its plugged. It was probably a hang on detection. Skipping detection by feeding appropriate arguments to ctmouse would have been my next step. I assume it is NOT a ps2 mouse, and is instead serial. Serial mice come in 2 flavors; ms mice, an Pc systems mice. The ctmouse driver assumes ms mouse flavor unless told otherwise.

I dunno. Ez drive always lives at the top of conventional anyway... With a bios that stays out of the adapter region, the best possible free umb is 128k. [128k for video mem area, and 128k for system bios are not negotiable.] In practice, its just 64k, because old games sometimes want ems, and you need a page frame. A 486 may be harder to get UMBs configured for inexpensively. I tend to use UMBPCI, and for obvious reasons that wont work here. There are some chipset specific utils for sis and opti chipsets that can do similar things though. I would try those... Soundblaster utils should never be loaded high. Ctcm and ctcu will do stupid things if you try. The big hogs are atapi drivers and mscdex. Emm386 is a bit fat at 4k use, but nothing compared to mscdex. I would use ctmouse for the mouse driver. It weighs in at 1k of use. I think i could get the 620k free without qemm. Again, a bit biased against it.

QEMM was never magic for me. Every time I encountered it, it was causing some kind of strange edge case to happen. Granted, I worked at a mom & pop repair place during that era, and so my sample is surely biased toward the negative. (people dont bring working systems in to be serviced....) Instead, I got to be very good at manually prodding DOS into efficiently using the upper memory area, and had a list of go-to DOS drivers that were both very compact, AND very reliable/capable. A copy of MSD to be able to peer into the memory map to see where things were loading, and how big they were, would let me do what Memmaker did, only more guided/fine controlled. The biggest thing, is to have a system with a BIOS that stays the hell out of the adapter rom region. Many bioses, especially in the late pentium era, wanted to jam all kinds of crap in there, like realmod USB drivers, PXE bioses that would NEVER get used, etc. To effectively use the adapter rom region to provide UMBs, you need to keep stuff out of there so you can do the remapping.

The issue I have had with SD->IDE adapters, is that they are almost always all based on the sintech CF->IDE, and that version of the firmware they have mangled multiple ways, simply cannot do 48bit LBA. This would be fine, if the adapters did not advertise (digitally) that they support it. This leads to all manner of problems when more modern OSes with 48bit LBA aware ATA stacks try to access large disks inserted. I have such an adapter, and I learned this lesson the hard way. It is vitally important in those chinesium SD->IDE devices that you use a drive smaller than 8gb. I am rather curious if your CF->IDE adapter is able to do 48bit LBA. *edit DERP, I am clearly very tired from staying up. CompactFlash is based on a subset of the ATA standard, and the LBA capabilities would be wholly dependent upon the actual module inside. (It is practically a straight-thru connection. For some reason this was just not clicking in my head-- I already knew this. Derp.) The issue with the SD->IDE adapters, is that there is a baked on SD->CF adapter bridge, then glued to a hacked CF->IDE adapter chip. It is this SD->CF conversion that is buggy. Later versions of discrete SD->CF converters (meant for high end digital cameras) can deal with large capacity SDCards, and do the 48bit LBA modes fine. I guess I should just stack some adapters, and buy a good quality CF->IDE like the startek one linked. *shrug*

(has a bunch of old CRT VGA monitors upstairs in a closet... Has considered selling them, but would need a local buyer. Decided to just keep them. They work fine.)

I am a tinkerer, and I recall a breakthrough moment I had working toward a now-silly goal. For reference, this was in the era when Blaster and CIH1019 were making their rounds. There was an outstanding need for rescue media, and people were still using win9x. I recall, that I learned that you can force a drivespace3 compressed volume file to get mounted using scandisk's DOS interface command line. (there is an undocumented /mount argument you can pass to it) This enabled me to use a more robust version of xmsdsk.sys to create a large fat16 volume in a ramdisk, and then compress it with drivespace3, copy the windows folder and other useful bits and bobs for it (such as a program files folder, and pals) inside, then save it on whatever media I wanted to preserve the image with, In this case, an el-torito booting CDrom. The El-Torito CD would load a floppy diskette image into RAM, and boot it. That image contained drvspace.bin, and other supporting programs-- along with a properly doctored MSDOS.SYS. It would also load an ATAPI cdrom driver, and the ramdisk driver. It would then copy the drvspace.000 file into the ramdisk, run attrib on it to make it writable, then force-mount it with scandisk. This created an "Uninfectable" recovery volume, which could boot to a full GUI, and thus run full versions of antivirus software. Later, I learned that I could further improve this fun little gadget, with ISOLINUX+Memdisk, which would make using the file copy, and force-mounting of the volume unnecessary steps. I was a very happy little nerd back when I got this all figured out. Cleaned a lot of infections with it. I believe I even used it for a short while after I had a hard disk failure, and had to save money for a replacement. (was rocking a properly configured liveCD image instead, until I could do the actual repair.)

There is also the "Odyssey client" I happen to have a copy... However, it is commercial software that was never opened by the OEM. Technically, the installer defaults to a "30 day enterprise evaluation" unless you present it with a valid license key. I dont have sufficient knowledge of the rules here to know if sharing it would be kosher. It's a bit of 'hard to find' software, all the same. It is a "Universal Wifi connectivity manager" that also happens to have support for WPA2. It does not care one iota about the chipset, as long as it has a properly configured 802.1a/b/g/n/ac ethernet driver loaded. Hang on, I will install it in the win98 VM, and take a picture. Of course, there are no 802.1x devices in my virtual machine-- but I can confirm it works with USB dongles, PCI cards, and just about anything else I have thrown at it.

I have been approaching this from the standpoint of: "Hey, we have this bit of <Name vintage manufacturing equipment> that runs on a custom interface board, that only works on Windows98. That machine **REALLY REALLY REALLY REALLY needs to be able to parse webpages with something more modern than IE6. The webpages will be wholly intranet, but the browser needs functions not present in ones from the era. Do you know how to make this happen?" Or something similar. This is why I worded my reply to CS1 the way I did-- "If you are unfortunate enough to find yourself in a position where you MUST do this..." If *I* were in such a position, I would also heavily leverage isolinux and memdisk. (because the win98 deployment will be HELLA unstable, and prone to eating itself.) I would create a drivespace3 compressed disk image of about 500mb in size, with windows, the installed browser, KernelEX, and any other "Mission critical" software, in a clean, healthy state-- put it in an EXT3 partition on an IDE->CF adapter, and have isolinux load it into memory and boot it using memdisk. The computer would need about 1gb or RAM to operate. (500 for the disk image in memory, and 500 for windows to function.) This way, it could eat itself every goddamn day-- and it would be a simple reboot to "Fix" it.

I have tried it exactly one time. I found that it made the install even more unstable than a win98 machine already is. I am familiar with KernelEX, and believe I properly captured the severity of how much it messes stuff up in my humorous writeup, but maybe I was mistaken... LOL

You will probably have to install cifs-utils. The linux parlance for "Really really old SMB" is "CIFS". Then, in smb.conf, change it so it has this line for the minimum version. server min protocol = NT1 Your Samba server will then be able to make shares the win95 machine can connect to. With cifs-utils installed, you can mount a share on the win95 host from the command line, so you can directly work with it on your linux box. sudo mount -t cifs -o username=<win_share_user> //WIN_SHARE_IP/<share_name> /mnt/win_share

That and others..... Anyhow, the "You should never ever do this, but if you are unlucky enough to find yourself in a situation where you MUST do it, here is how" for win98 is complete, and a thread made. I will do the "Much more useful, you might actually use this, and it will make you happy" thread about using the win2k Services for Macintosh will come another day. I will have to dust off a linux copy of Basilisk II and get some classic mac OSes downloaded to do that one.

By request, here is a nice graphical how-to on how to do something you should never do. Namely, getting an old windows 98 system up and talking on the internet. (Because it looks awful to use 16 color VGA, I have installed a generic VESA based display driver for use with this VirtualBox VM. If you are interested in it, it is an old project called VBEMP, and you can find it here:) https://bearwindows.zcm.com.au/vbe9x.htm#2 (It is for people who either have a card that has a Vesa Bios but no windows (9x) drivers, or for situations where you need a universal display driver that does more than 16 colors. It could conceivably be paired with the free Scitech Display Doctor. It does not do windows accelleration, has no 3D support whatsoever, and its only claim to fame is that it will give you high color modes on generic SVGA hardware, using the VESA bios in the card/TSR.) Straight from a fresh install, aside from the above (and pushed to 32bit color!), we have our lovely Windows98 second edition desktop. Straight away, it looks like the installer successfully found the emulated AMD PCNet family ethernet card, and installed TCP/IP and pals, automagically. Since TCP networking is already configured and working (I used my NAS to get the VBE driver into the vm! ) Let's just focus on the needful to get this thing sorta half-assed on the internet. First things first, we install IE 6 service pack 1. You can get it from Oldversion.com (because Microsoft does not have it available anymore) here: http://www.oldversion.com/windows/internet-explorer-6-0-full-installer This is the last "Officially supported" version of internet explorer for win9x. Before we even try touching the internet, we really kinda need this. While I question the validity of the user agreement in this modern age, you must agree to it to proceed regardless. Since this is the "Full, offline installer" package, you really should do an "Install minimal or custom" type install, because IE will **NOT** be able to pull the missing features later, should you encounter a situation where it is needed. It's better to just have them all installed right now. Then go through the laborious process of enabling all the components for installation... This is a Post-Flash world now, please, PLEASE, PLEASE, DO NOT INSTALL FLASH PLAYER. You HAVE been warned, Thank you. Things you SHOULD install are DHTML binding, Additional web fonts, Media Player and its codecs, etc... The rest of the setup process is pretty straight forward, and when done, will ask you to reboot. When we first start IE after installation, we will be greeted with Microsoft's Ill-Fated attempts at being an ISP... We have to placate it a little, before we can proceed... If you did not install Outlook Express during windows setup, it will now force you to install it, because *YAY VERTICAL INTEGRATION!*-- Who doesn't want the most insecure email client ever installed on their vintage computer? Well-- Microsoft doesn't give you a choice here. If you already have it installed, you get asked what kind of modem you intend to use. Yes, I know what you are thinking--- "Wierd, this is madness, I don't intend to use a modem at all!" well, yes, we are getting to that. First though, Microsoft is REAL INSISTENT on being your ISP, and they thought you would be using a dialup modem, and you will jump through the hoops like a good user. Pick "Dont detect, I will select my modem." It profits nobody to search for a modem that does not exist. For now, we have to nod our heads and say "Uh Huh." to the setup wizard, before we can get to the options to work with a LAN. Since we are just trying to get past the "Join MSN Online! Wow! Best Internet! You'll See!!" aggressive marketing, we have to just roll with it for now, and pretend to be a rube, ready and waiting to end up being billed monthly by Redmond.. We do that by saying we have a totally fictitious modem on com1. We will remove it immediately after we get past the forced attempt to get us to set up their unholy internet service. After it sets up a driver for our totally fictitious modem, we **FINALLY** get the option to go "Hey, Microsoft-- Stop the BS, I use a LAN!" Now that we have it, we click the button. LIKE MAGIC, the "USE OUR INTERNET SERVICE, PLZ PLZ PLZ!!" garbage just goes away, and you get a much less abusive setup process. It might be kind of confusing, to have to tell the setup process that you have a modem, and pretend to set up MSN Online just to get to this point, but hey, there is a REASON Microsoft was subjected to antitrust investigation in the 90s.... Moving on. Unless you TRULY want to live like it is 1998 all over again, you do not need to mess with *ANY* proxy server crap. Just click next. And just like that, we are all done! Now, we should remove that stupid fictitious modem driver.... ahhh.... Much better! Now, let's see if it actually works... Looks like it does-- but wait--- what's that? No HTTPS in the URL bar!? Well... I wonder what happens if I... Oh NOEZ! Well, we dont really want to use IE6 to go on the internet anyway---- We want something a little bit newer. And for that, we will need a totally nonstandard, buggy, system destabilizing, and generally not a good thing to do modification for windows 98, called... KernelEX http://kernelex.sourceforge.net/ Last updated in 2011, it's a mere 11 years old, instead of 30 years old. YAY! Let's install it, and hope it doesn't crash our VM in spectacular, new, fun, and amazing ways! After being greeted by a very familiar looking install wizard landing screen and clicking next, we get the GPL license window. Finally, a License I can agree with! (and actually mean it!) then we get told that our version of windows doesn't have unicode support. Why? Because 1998 was a simpler time, when classic computers still roamed wild and free, that's why! we DEFINITELY need Unicode support to use basically any modern browser, so when it asks, YES, We do in fact want to install that missing support. (you can install it yourself if you hunt down the unicows update for win98) Since that file is no longer on microsoft's servers, we can't have the installer pull it for us. We have to bail, and install the unicows.exe update ourselves first. Because who wants to support old OSes that only old neckbeards in basements want to use? Thankfully, for us-- Archive.org has our backs. So, let's bail on the installer, then install unicows update. There, that should make KernelEX's installer happy... YAY!!! Yes, we do want it to work for all applications... so we click next... and then we reboot... and hope it doesn't bluescreen... AND, we have our first problem with KernelEX... Apparently, it DOES NOT like the universal generic vesa driver! what a lovely black screen! Some prodding in safemode to remove the driver and go back to ugly 16 color mode, and we are off to the races... Now.. Let's Grab Firefox 8 from Oldversion... and then install it... But OH NOEZ!! WHat haPeN!? Well, if you read the readme for KernelEX, you will see that they added a compatibility tab in the shell user space, so that you can assign a specific windows version lie, among other things, to help get software meant for.. ahem.. Newer versions of windows... to run. right click the executable, choose properties... then the KernelEX tab, and then tell it that you want to pretend to be win2k SP4. YAAAAAAY!!! But does it actually, you know--- WORK? After adding an exception (because it cannot verify the modern digital signatures...) It seems to be! Look at that, HTTPS, on a modern, live site, From win98se! Now, mozilla arent dummies-- they really want you to upgrade that browser... Let's see how far KernelEX will really take us. KernelEX hasn't been updated since 2011 after all... So we shouldn't really push our luck here, but let's try some newer versions... Looks like we can push it up to firefox 10 ESR. It needs an XP Service Pack 2 lie with KernelEX to install and run. While not the.... Greatest security level.... We can at least poke around a LITTLE bit on the modern web. Youtube of course, tells us to go get bent, but whatever. At least we can display the page telling us to go get bent! The oldest version of firefox that supports TLS1.2 is Firefox 28, and it does not like KernelEX. (rather, MSVCRUN100 is missing necessary exports, and since KernelEX is not getting patched anymore, you wont get it.) This means this is about as far as you can push your luck with getting an ancient win98 machine onto the web.

That has been the consensus. Yes. There might be reasons for wanting it though. (Ancient game that wont run in (modern) windows or WINE, that does online play with dedicated server, for instance) Most of the security issues come from the SMB stack. If it is not bound to TCP/IP (and uses IPX/SPX instead), it would be much more safe. Not "safe", but much "more safe". (Winsock2 has its own fair share of critical vulns...) Despite making a graphical how-to, I think I have peppered it with enough "Subtle reminders" that this is not a good idea.

OK, I will make some fun threads.

From what I am reading, "yes", with scarequotes. It has a form of MS Basic on it, but not a lot of memory free afterward. No storage medium to save the programs to. The system bombed bigtime at market, and never got peripherals. I do not know if there is an enthusiast support group or not-- If so, it would be your only real bet at getting it into an actually useful microcomputer state. In theory, yes, you could learn MS BASIC with it. That would teach you about loops, conditionals, operators, and other components of writing software. There might be some modest crossover with MS VisualBasic (and its variations in .net) that might be somewhat useful today. I would suggest a different micro to do that with though. Say, A C64.

Did you guys want me to do another one of these for win98, or for the more exotic, "Services for Macintosh" using a win2k box? I would start actual threads for those, if there is interest. You can actually get a surprisingly new browser on win98, using the Kernelex package. (patched kernel32.dll, that includes functions found in win2k and xp, giving access to some reasonably newerish firefox builds.)

Now that I am home, I can confirm that my consumer grade NAS box simply just works. (It comes pre-configured to talk with old SMB1.0 hosts.) But what about if we want to mount it as a drive letter, and then reconnect at logon? Well--- Right click on the shared folder, and choose "Map Network Drive" A little dialog box asking for a drive letter, with a "Reconnect at logon" checkbox on it appears... Click the checkbox, set a drive letter, and then click OK. You will then have a new drive letter in explorer, that is associated with that share. This is why I suggest having a consumer NAS box. It makes this MUUUUUCH easier. Otherwise, you have to jump through the hoops I linked to earlier, on your modern windows box, to set up a suitably unrestricted fileshare using the old protocols. But suppose you want to go the other way, and share the win95 system's drive, so that the modern machine can see it-- Well, that's doable too! First, right click on the drive or folder that you wish to share, and choose "Sharing" A new window pops up, asking how we want to control access. It looks like this when it first opens. We want to share it-- So, click the Shared As radio button, then populate the info for how we want it to appear on the network. Say, something like this: We are giving full access, so that our main machine can write all over it remotely-- Click OK. As soon as you do, the drive or folder you set up sharing on, will have a little blue sleeved hand underneath it, like this: and our host OS of choice should be able to see it also-- like this:

PART II --- With IE 5.5 SP1 installed, we can kinda sorta see google.com now! YAY! However, in my forays into getting a newer browser installed, I have hit several walls with win95. Basically, everyone dropped it like a stone when 98 hit the scene. In a few hours, I will demonstrate file shares with win95 all the same. Give a bit.