Take DOS, stir in some Netware, add a bit of Windows and... it's ALIIIIVE! As the weekend recedes, be glad you didn't find yourself in need of a fresh change of clothes like the contributor of today's entry into the Who, Me? hall of shame. Today's story comes from a reader the wondrous Reg name maker-upper has elected to call "Miffy" and concerns some Teutonic exploits that left our reader tired, …
Miffy managed to restore some of my long repressed memories. Thinking of it, I still wonder how I ever got some Frankenserver-client-network working in the mid 90s. And if that wasn't dodgy enough, we also had some Macintoshs with PC cards running Windows that were expected to all work together. The horror! the horror!
This post has been deleted by its author
Ahhh...QEMM...I remember it....not Quite a Memory Manager.
I had a job doing schematic design on a 486. To run the schematic capture software package, you needed to load QEMM. All went swimmingly, until you tried to open one too many schematic pages, or put too much circuitry on one page.
At which point, QEMM threw up its hands and fell over, taking Win3.1 and DOS with it.
My solution was to install Linux (ver 0.9-someting, 13x 3.5" floppies IIRC) on a second hard drive and dual boot to the Linux drive on the same hardware. I ran the Linux system (same hardware, mind you) as an X-client off a Unix system that had the "better" version of the schematic tool (Workview?) and all was well. Better than well, actually. Performance was better, reliability was great, and I never used the Windows version of the software again.
It was at this point that I realised Windows was, to put it politely, not making the best use of the available resources. Whereas this upstart Linux seemed to be doing exactly that, and providing impressive performance to boot. I continue to hold that opinion.
The supposed "640K limit" was an IBM hardware limit, not an MS software limit. The IBM hardware spec was already firmly in place before Gates even heard about the project, so even if he had made the comment (which is extremely doubtful), he would have just been agreeing with IBM. And it wasn't really 640K, it was more like 704K, if you knew what you were doing. I find it absolutely amazing that this piece of incorrect trivia is still being parroted as fact after all these years ...
OTOH, I personally remember Steve Jobs saying that "128K ought to be more than enough for home users", at a meeting of the Homebrew Computer Club in late 1983, as he was demonstrating the original Macintosh, just before the public unveiling. At the time, he had a point ... people were running flight simulators in 64K!
On the gripping hand, none of this matters any more. It's all just an accident of history.
Ah, but wumpus ran in 1k on the KIM 1.
It cheated, though, by outsourcing graphics to your head - a hex display is not quite the same.
Still, it was fun. The first time I got into some graphics game was with Doom, which was immediately considered an essential Friday afternoon "network test", and I recall we had to do something profoundly filthy with Netware to get it to work amongst ourselves :)
Or the limits of the chip and systems of the era.
The PC was started being designed in an era when 4k was in common use, and was released with 16k as the entry level point when the hardware was ready. It used the 8088 chip, which could address a max of 1MB, and IBM needed room to map the BIOS, video, etc. into the upper areas of the memory map, which was common in that era before virtual memory mapping chips became defacto. That is the 640k dividing line, at a point of 40 times the memory that the base unit shipped at. Even if they had it higher, the BIOS and slots still needed to be mapped in somewhere in the 1MB space that was the max the 8088 could address.
The Apple II had similar design constraints from an earlier time, but nobody goes on about the 48k limit of its design until they did some bank switching on the ][+ to get that extra 16kb of RAM or more on the //e and //c.
XMS 2.0 wasn't done until 1988, (seven years after the PC was released).
Even when PCs were shipping with 80286's, which could address a whopping 16MB of RAM, most PCs of the time didn't have more than 1MB of memory.
By the time the 80386 came out, and things could address lots of memory, it was the software lagging behind, with most people insisting on running MS-DOS of the era of 16k machines.
Or the limits of the pocketbooks of the era!
I laugh at hardware reviewers today that act like a 20 buck price difference is going to mean something to most of us who started back in the DOS and earlier days. When I got my first 486 - I spent $350 on 8 MEGS of ram, and $60 on a half meg of vid ram (so I could run real svga). And I still remember when someone came into the shop and ordered 128 megs of ram for the giant database he had to run. We were in awe.
I have a copy of an invoice for an 18Meg drive that set my customer back $4,200 in July of 1980. It was a North Star HD-18, plugged into a parallel port on a North Star Horizon to supplement the overloaded two year old stock 5Meg drive. The system ran a proprietary, home-built inventory and invoicing system for a local indy auto parts store in Mountain View, California. A guy from North Star arrived with the unit to swap out firmware, update the OS, and make other changes so the machine would accept the second drive ... there was no charge for his services, including travel from Berkeley. It worked quite well for about a decade, when I upgraded them to a Coherent based system, which was followed by a Slackware system about 10 years later.
I remember seeing an electronics magazine advert from the 60s selling core store memory at the incredibly cheap, according to the advert, price of 3d per bit! (youngsters 3d = 1.25p, then inflation). By my calculation that comes to over £100k for 1MB of ram - and that is 1960s £100k. I started work in the 60s for £5/week. That's Moore's law for you.
My first experience of buying ram was IIRC buying 12k ram for an Acorn Atom for £50 in 1980
I read somewhere that it was for backward compatibility and virtualisation.
The idea was that you would have your CPM programs running in their own 64K space, totally isolated from each other. That's why you had that weird memory addressing page:offset scheme.
10x640K = 640K, ten virtual machines.
VMWare - eat ya heart out!
That "weird" page-and-offset memory addressing was because the 8088, like the 8086, had a 20-bit address space but 16-bit registers. It was influenced (according to the iAPX 86 designers and others) by a desire to ease porting from 8-bit CPUs such as the 8008, 8080, and Z80; but in reality 16-bit registers were probably as large as was economically feasible.
I doubt it had anything to do with CP/M, aside from the observation that CP/M and applications for it would be numbered among the software packages Intel would expect people to port. They were primarily thinking at the instruction-set level, though; while they made some concessions for compiler developers, I doubt the designers spent much time considering specific applications.
It's worth noting that the 8086 was developed in part as a stopgap while Intel was working on the iAPX 432, development of which started in 1975. The 432 had a 32-bit address space but still used 16-bit registers because going all 32-bit at the microarchitectural level just wasn't feasible. Considering how ambitious the 432 was in other ways (a capability architecture implementing a stack machine with an extremely CISCy instruction set), that's pretty telling.
If you want to blame anyone for the limit, blame Intel. Had they used page aligned (256B) segments rather than paragraph aligned (16B) ones for the 808x, we would have had a 16GB address space for Real Mode programs rather than a 1MB one. The hack of banking switching memory above 1MB into the HMA would have been totally unnecessary.
If you want to blame anyone for the limit, blame Intel. Had they used page aligned (256B) segments rather than paragraph aligned (16B) ones for the 808x, we would have had a 16GB address space for Real Mode programs rather than a 1MB one. The hack of banking switching memory above 1MB into the HMA would have been totally unnecessary.
I doubt any one could've afforded it. I remember in the mid-late 90s 2nd hand 72pin RAM was >$NZ50/MB. The machine could've addressed more, but my wallet wouldn't!
> I doubt any one could've afforded it. I remember in the mid-late 90s 2nd hand 72pin RAM was >$NZ50/MB. The machine could've addressed more, but my wallet wouldn't!
I had 1 MB in my Amiga 500 back in '88, as did most of my friends. By the early '90s, most of us had 2 to 4 MB in our PCs, STs, and Amigas.
I remember when Mechwarrior 2 was released in '95 because it required at least 8 MB of memory and I only had 4 MB in my PC. It was an expensive upgrade, but it wasn't prohibitive.
Ah - QEMM386 - great program.
I originally had a 286 with a 2Mb Expanded Memory card (using ZIP memory), using the ISA slot and running SmartWare and "As-Easy-As" (a Lotus 1-2-3 "clone") I could load huge spreadsheets that on PCs with standard 640k RAM, would just keeping paging to disk.
Then upgraded to a 486-33 with 16Mb RAM and it was like going from a mini to a Rolls Royce....so quick and effortless to have so much RAM....and QEMM386 kept it all running, no matter what I chucked at it.
p.s. Remember this?https://en.wikipedia.org/wiki/DESQview#DESQview_and_QEMM
Yup.. Loved it, and ran it on my BBS machine (which was also my only machine for a while). 2 nodes + one for me, as well as some extras.
There was a 'spawn" or similar named program for Desqview, and this would open a new 'task window" (I'll call it that coz I don't recall the name) and run whichever command you'd fed it. For the BBS, when it received mail from another system, I made great use of "spawn" to get the processing happing in a "background" task, allowing the front end to recycle in a few seconds (Binkleyterm IIRC, though I played with Front Door, Xenia and something else a few times. Then again, maybe the BBS ended on Xenia.. Was some 20 years back :( )
ah - 4DOS - what a great replacement for MS-DOS command.com
I can even recall the wonder from my ex-boss when I did a "dir/w" on a PC and I'd set specific extensions to be displayed in specific colours (which 4DOS allowed) on the EVGA monitor I used.
He couldn't believe it, as he was used to white text on black screen when using DOS.
And for real responsiveness, did you also make a 16kb ramdisk in upper memory and put copies of all your favourite .bat files into it? I know I did. People were shocked when I could just type one command and 'immediately' I'd be in the right working directory and the program would be starting up.
We had DOS clients, Mac clients (Mac Plus and Mac SE) using AppleTalk, a PC with an AppleTalk to Ethernet (thin cable coax) and a VMWare server.
Stable as houses. Well, until the training manager decided to take his PC with him to do a training course and simply unattached the coax from either site of the T-connector...
I pushed my team through CNE training and certification. It was all paper based until the exam. There were a couple of errors in the answers one of which would make you fail. Candidates were briefed on the 'acceptable' answer before the course, on the course and before the exam, I still had one very able technician who could not put the 'wrong' answer into the exam and took 3 attempts to pass. He had a PHD in nuclear physics.
Anybody remember the name of the nlm on Netware 3 that repaired volumes? Remember it being the thing I copied to the DOS disk after setup cos if you volume didn't mount you needed it, and it was on volume after install...
Cannot remember its name though. A quick search says vrepair.nlm or toolbox.nlm for later Netware but neither sound right for Netware 3.
Sounds like my then employers migration from NetWare and Win 3.11 to NT4 on the file/print servers and Win 95 on the desktop. A flleet of machines cobbled together from random parts had to be upgraded one by one in order to support Win 95. Thankfully the dev and production systems continued to run a mix of VMS and Solaris, but outside our dev department it was chaos for months.
"outside our dev department it was chaos for months"
Just months? For most places, it was years. Decades, in fact. Still is.
How many man-hours world-wide would have been saved if the corporate world had sensibly told Redmond to fuck off back then? More to the point, why the fuck do people still allow the garbage from Microsoft on their corporate systems? I wonder how many billions of dollars (trillions?) have been wasted in this charade?
Bite me once, your fault. Bite me twice, our fault. Bite me thrice, my fault. There will be no 4th bite, and if I have anything to say about it, no third either.
Because they provided something that appeared better than before, and was "good enough" for most people to accept the shortcomings. Then MS got into some of it's dirty tricks, and by the time people realised what was going on, MS had managed to entrap users with proprietary everything and killed off the competition.
There is a reason why I look like this >>=======>
I still bear the scars of PC/MS-DOS V1.x (on 5150s), 2, 3, 4, 5, 6 & 7; Windows 286, 3, 3.1, 3.11, 95, 98, etc., running on NetWare/IBM/DECnet/*NIX networks.
I think the market decided it was "just good enough" or "nearly good enough" and that that was "good enough" "for now", and that MS would fix it and make it better.
Windows 3.51 was OK, but NeWare was better. Unfortunately, I won a bet with an OEM that Windows would replace NetWare (because it meant that people would think it was all "Windows" and they did not have to support two systems) even though we both thought that NetWare was "much better".
Programme - programme for anything except computer. Though what about linear program(me) - I remember a university exam paper saying programme, but maybe that is a bit eccentric?
What about dialog though? Do program dialogs follow the same convention of always being dialogs, or not?
Even in 2008 a migration from NetWare to Windows servers could be iffy. The company I worked for opted to use a consultancy to design the Active Directory database and they decided to do things very differently to how it had been done with eDirectory. The end result was that it took the PCs four minutes to execute the huge login script and a rights issue that allowed every client to see each other's supposedly private home directory. The old eDirectory login script had a sub-second processing time and there had been no rights issues at all. On top of that the budget for the servers was so low that the backup took about twenty hours. The servers had to be replaced in less than a year...
I did it in 2013 - moved from Netware/SLES/eDirectory to Win2012. It wasn't too bad with 2012 - since Powershell is a first-class component, not an add-on, the scripting is scads easier than using batch files. It did take a lot of little scripts to create accounts, move files/etc over, set the permissions. Then there's the whole learning-process of doing drive mappings via GPO instead of centralized login script. Personally, I still think Netware's client login script processing is worlds ahead of the GPO garbage, but the world moves on. OK, full confession - I still think eDirectory is a far better directory service than AD, and Netware file serving (and being able to give permissions to OUs instead of only groups) beats Microsoft's offering, hands-down. But, like I said, the world moves on.
The problem was that Novell Netware ran like a demon and Microsoft replacment didn't. You couldn't set up a Netware server unless there was enough RAM to cache the bitmap of the allocation map for the drive.
Also, Netware didn't waste CPU cycles on a graphical interface, so the time to transfer data from the disk to the network card was always as fast as it could be.
Any objectve comparison between a Netware network and a Microsoft one (using a Network Analyser) would always show that the MS one was ... poor.
But poor old Novell didn't really get that Microsoft were activily working to kill them off.
Oh dammit.. repressed memories time... Netware 6 (?) introduced an absolutely appalling GUI that was unstable as hell, slow as a superglued frozen slug and only had half the functions of the old text based UI. Which helpfully had some of the functions removed and shoved into the new wonder-GUI.
Yeah, Novell's port of xwindows to Netware NLM was hideous. Seems like the only time I remember using it was something having to do with Groupwise upgrades, or maybe SSL certs. Otherwise, avoided it like the plague that it was. Then they went to Suse and decided that monitoring screens, and even the half-baked xwindows screens, were giving out way too much information, so they stopped showing ANYTHING AT ALL on the terminal and shoved it all into log files instead. That was, IMHO, the beginning of the end of my love affair with Netware. And it didn't help that dsrepair under SLES needed a bazillion command line switches to run instead of having a menu system like <deity> intended.
Yeah, it wasn't just Microsoft that killed-off Netware, Novell did a pretty good job of it too, there at the end.
"The wonder-OS [NT4] needed at least 16MB (preferably a bit more) and a decent CPU while Windows 3.1 would toddle along with 1 or 2MB and a considerably lower class of CPU."
When I first started this job, those were the specs of the PC I was assigned. My current work laptop has 16GB RAM and a 6-core i7 with Windows 10 as the OS, all to run PuTTY.
You had a desk to sleep under? We would have LOVED to have had a desk to sleep under. We had to sleep under soakin' wet Lotus 1-2-3 manuals, out on a concrete floor, and if we were REALLY lucky, the rats wouldn't run off with half the beads on our abacus, so we could keep on doing calculation when we had to start work at 3 in the mornin'
I remember 20 odd years ago asking why there were no drivers for an HP scanner for HP machines running NT (I was working for HP at the time). "HP's view is that Windows NT is the future of personal operating systems, and the company is determined to keep it there" I was told (and he kept a straight face...)
While the reasoning seems unsound, the decision not to develop an NT driver wasn't. Even back then, i.e. NT 4.0 era, I couldn't believe that a considerable part of writing an NT device driver was trial & error.
It might very well be that I was simply crap at developing DDs. And I never touched anything like that ever after.
Last week I was trying to get an old HP printer to work on a PC that was recently 'upgraded' from W7 to W10. The troubleshooting guide recommended downloading new HP driver and support software.
278 fscking MB!!! For a printer!!
Reasonable printer hardware, but HP still can't write software worth a damn.
The last time I had an HP printer ( or indeed anything else) it refused to update the drivers.
The update fell over at one particular DLL that wouldn't delete/overwrite/replace. So the update stopped .. So no printer.
Manually uninstalling the old driver wouldn't work. Despite going through the various depths of HP's uninstall software.
The specific DLL couldn't be manually uninstalled or deleted. Whatever HP had done had fixed that DLL into place stronger than concrete. Nothing would shift it. And the install software wouldn't ignore it. It just aborted installation Which was really the last straw - because the new drivers used exactly the same fucking version!!!!!!!!!!
I junked the printer and have never bought anything HP ever since.
The specific DLL couldn't be manually uninstalled or deleted. Whatever HP had done had fixed that DLL into place stronger than concrete. Nothing would shift it.
Whenever I came across an "immovable file" on a client's machine, I fed it some Mint. Seemed to clear up whatever was ailing it right quick!
(Usually took an image of the disk first, but not always - depended on the value of data/whether the client was certain they had a recent backup etc)
That's the "luser driver" offered to the unaware user which will install all those HP software for school projects, direct cat photo printing, bundled third party offers, cloud printing, and instant supplies ordering without your consent anytime the printer feels lonely.
If you look, there's probably a simpler installer with only the required files, and sometimes even a zip without the installer (to be used with the Windows "add printer" tool). Many printers are also supported by the HP universal driver.
OSX upgrades had the vendors abandond almost every scanner made for OSX in the past.
My Fuji Scansnap used to bring me smiles using it about just how well it ran, which changed overnight to one of the most frustrating things I own after OSX got upgraded.
VueScan can handle the older hardware when the makers all abandoned it. I'm not sure its at a level of bringing me smiles, but at least it is functional.
Not in the same league but in around 2008 I was consulting for a few days for my recently former employers. The new PFY desperately needed to get some data off an old DOS machine, used for a weighing system, that didn't have working USB or even an FD. It did, however, have a network card (NE-2000 based of course).
Said PFY didn't realise you could run up the network and connect to the server using only DOS commands.
"The world," recalled Miffy, "looked decidedly rosy, which might have been why my then employer let my 25-year-old neophyte self be the technical lead for our migration to NT4."
Maybe her employers had been celebrating a little to hard and too long which gave them a decidedly rosy view of the ease of network migration? Having said that I well remember that at 25 I was also a stupendously experienced, fantastically capable engineer as well......
1994, MS Central ...
MS1: "These NetWare guyz, how can we get rid of them ?"
MS2: "Act I, get them to run as an application on our server, that way we get the licenses for the server. We offer unlimited # of users to connect to NT workstation print servers to make up for the price diff."
MS1: "Nobody would be that daft, who would trust Windows to be as stable as NetWare. Also, what is the point of NT Worstation with unlimited connections, we could monetize that."
MS2: :"That is the point, along the way, Act II, via updates, we kill some printer driver, but only if NetWare services is running. Then they switch to us. As for NT Workstation, we first need to kill NetWare, then we can milk people ten times more ... as we'll have no competition"
MS1: "Son of a batch job! We'll never get away with this!"
MS2: "Worse, Act III, we will have them connect to our servers for licenses, to make sure they all pay through their noses."
MS1: "Naah, dreamer"
MS2: "Then, we hook the license to hardware, when a part changes, they'll need a new license from us."
MS1: "..."
They still are getting away with this!
Windows server software required CAL licenses from the start. The big difference between Netware and MS is MS didn't _enforce_ license count, whereas Netware did.
After years of MS making it near impossible to actually figure out how to properly count, license or even buy said CAL licenses to be properly licensed, just about every windows user punted and didn't care.
Most people buying file servers compared the properly licensed cost from Netware against a vaguely licensed MS Server, or perhaps CALs didn't even come up in the conversation. Once the rollout went out and they weren't counted, audited, or were anywhere to be found, it was instant conversion over.
Hah... Since Novell was so finicky with CAL's...
If you connect DOS workstations to a Novell server, each DOS workstation takes one CAL.
But if you connect a Windows wankstation to a Novell server, it'll happily use two (or more) CAL's... which will give you grief if you have 9 dos users, 1 windoze user and a 10-user Novell licence...
...it happened to me, I now got severe flashbacks, double damn.
Having period flashbacks here.
For those of us on the Mac side of things, mid to late 90's networking brought us the wonders of Appleshare, the glory of Open Transport and if you wanted to pollute the network with Windows then of course you needed Dave. Not a bloke called Dave, there really was SMB networking software called Dave.
Functionality for all this was of course at times optional.
Extension Manager was your best and only friend.
This post has been deleted by its author
Netware 5 and friends.
Customers were not allowed Windows servers until.
1) Our database could run on it.
2) Our customers had hardware support.
3) Server 2003.
4) They accepted the performance decrease.
We did have one NT family to Netware upgrade as they were hammering the server too hard.
It was always inevitable that as computers got more powerful / cheaper that Netware would be pushed aside by something that didn't need the Netware priesthood with their NCSOE (Network certified something or other) qualifications. They had a good run for their money, but if you make yourself the priesthood of something that is clearly going to be commodified you are going to end up as roadkill on the information superhighway.
I am slightly bitter as we had one who not to be shortist or beardist or gitist was a short bearded git who was a neighbour and despite being a colleague complained to the council when our burglar alarm malfunctioned.
The main thing of course was that to run anything else on the servers you needed a NLM (Netware Loadable Module aka No Laughing Matter) with a complex API and no support for normal C language runtime APIs. So no one wrote them really. Unless I am imagining it you needed a very expensive version of the Watcom compiler to build them? (Also wasn't it easy for them to bring the server down - certainly that was the case with Netware 2 in which they were called something else daft, was that fixed in 3 or 4).
@John_Styles I think you are recalling VAPs - or Value-Added Processes as they were called, which predate even my experience of NetWare.
They - to quote from the Novell site - "provided a limited interface and an inflexible environment, forcing supervisors to bring down a server and then bring it back up again to load a VAP."
Ah yes VAP.
Do you think this is right? https://en.wikipedia.org/wiki/NetWare_Loadable_Module about NLMs being in the same address space in NetWare 3? I thought that was VAPs in 2 and was changed for 3 (we didn't actually write one but we thought very seriously about it and bought books and stuff)
They were certainly cooperatively multitasked (rather than preemptively) so when someone installed a SQLAnywhere module on our file server everything ground to a halt. I also had personal experience of watching the NetWare server ABEND from a faulting NLM and demand a floppy disk to dump the memory to.
The following link took me so far down memory lane I was humming Dubstar songs:
https://support.novell.com/techcenter/articles/ana19910901.html#d12jhwc
...but the TCP/IP stack stubbornly failed to LOADHIGH resulting in even crapper PC performance."
There was a hack for that sort of thing that I used to dine out on at the time. Chuck whatever monolithic pile of shite ${network} had foisted on you as a client in the bin. Chuck on the Novell card driver for your hardware and the link support layer from the ODI stack. Then load the Novell client for your network onto the LSL[1]. Even if the (now much smaller) client wouldn't LOADHIGH, the fact that the heavy lifting bit would took the edge off.
Also the Novell drivers and LSL were bloody bulletproof, whereas whatever ${third_party} had provided to talk to the wire invariably wasn't. If you were running more than one network protocol this was a "must have" solution as the alternative was a teetering pile of shims on one of the aforementioned shite monoliths and a "one down, all down" effect.
[1] In this particular case, I have to suspect that Novell's TCP/IP implementation might have been a better call.
Nijntje, diminutive for konijntje (little rabbit). But yes, Miffy for the rest of the world. There's a statue of Nintje on the Nijntje Pleintje, near where I used to live in Utrecht. :)
[edit] fine, search result: https://www.holland.com/global/tourism/destinations/utrecht/nijntje-pleintje-1.htm
"I have yet to see this on any Windows server, unless you use low-level disk editing tools."
Look harder then.
Shadow Copies provide the same functionality. Introduced in 2003 server.
Monitoring file access (auditing in MS parlance) has been there since NT 3.1
It was the 16 bit version, and I had to edit it so it just booted Windows and didn't run anything else at the start but somehow it worked.
Until you tried to run anything more heavy than notepad then it either crashed or slowed down so much you had to reboot.
I have up and downgraded to old Dos 5.x and Windows 3.1, that worked way better.
I also made a machine dual boot Freedos 1.0 and Windows 98SE thanks to old Grub 1.x, yes the one Puppy Linux still uses, but that barely counts as anything worth mentioning.
It was the default version that was for sale. Why do you think thar it booted perfectly fine if you deleted the WIN32 directory? And no problem if you had to delete it due to a virus. Most CDs from PC magazines back then included the 32 bit library as an utility to install and run orograms, shareware and demos in the CD that needed them. Then you just had to install again any program that used a Win32 file.
OEM version 2 and later (Not for same but included preinstalled on PCs) included a better integration of the WIN32 libraries but it was still a 16 bit system also running 32 bit libraries.
That's why Windows 95 could run any program from older versions of Windows.
It took until Windows XP for Windows to be unable to boot the graphic GUI without the WIN32 directory.
"It was the default version that was for sale. Why do you think thar it booted perfectly fine if you deleted the WIN32 directory?"
Windows 95 didn't have a WIN32 directory. Nor was there a SYSTEM32 folder - that came with NT and its derivates.
Whether you delete some files and folders may or may not hinder the boot process, and proves nothing.
Running Win95 on 16-bit hardware would be pretty extraordinary feat since so far it hasn't been done by anyone yet.
I think you need to show some evidence now.
"OEM version 2 and later (Not for same but included preinstalled on PCs) included a better integration of the WIN32 libraries but it was still a 16 bit system also running 32 bit libraries."
No-one is disputing the fact that Windows 95 still contained 16-bit code. It however had a 32-bit kernel with plenty of 16-bit code included as well. There are probably plenty of books and old articles about Win95 online, check Wikipedia first.
"That's why Windows 95 could run any program from older versions of Windows."
Yes. There were compatibility layers. It could even run DOS software. So?
Been through most of that.
It started in the 90's when I worked at ACT who had been Apricot Computers (anyone remember them?) and had a green screen terminal to work on an application that had colours for indicators, not the best idea.
So I trudged to the basement full of old PC parts from the Apricot days, they had proprietory memory chips etc... which was fun, cobbled together some working machines then managed to get Windows 3.11 working and even on the network. Things moved on from there, I wasn't even an admin back then just a junior support programmer.
In 1991 I was hired by a certain Blue automotive company in the USA to build client configurations using QEMM memory manager to support WIndows 3.1x clients with the PC/TCP TCP stack and Hummingbird Xceed X-windows clients for their engineering campus in a solution that used Unix boxen as file/print/application servers (because Windows Server didn't yet exist and this automotive company despised Novell). We also decided to give all the clients access to this glorious new fangled thing called "the Internet" by deploying the NCSA Mosaic web browsers as a standard part of the client build. As this was the birth of web browsers, we also had to compose this magic file called the "mime types" that would perform background hocus pocus of actually starting the correct application and displaying the double click selected file from a share or local drive based upon the files extension (.doc = word, .xls = Excel, .mif, .fm = FrameMaker, .htm,.html = Mosaic, etc.).
We had one user migrated to the new PC/server solution that had been using Word Perfect and Lotus 123 for DOS before being migrated and took it upon himself to be clever and rename all of his .wp wordPerfect files and all of his .123 files to .doc and .xls thinking that the magic would transform his old files to their brave new world with equal sorcery. Several frantic days of trying to figure out "why did the migration corrupt all of his word documents and spreadsheets" and a beautiful batch file written to search for all .123 and .wp files and attempt a automated backup & conversion as the machine came up on first boot and connect to the network righted the ship and corrected course. Of course the conversion executable only ran in dos mode prior to windows startup which caused an hour or more first boot delay.
Sorta doing that right now, if you count virtualization...
Old work laptop was a win7 machine with many moons before mine, a bunch of baggage, and arcane installed software to support legacy stuff. This version, before that one, but only after this other one, patch this, do that... And that's just the one vendors stuff.
New hotness Xeon hexcore lappy with win10 and bagonza ram pops in over holiday break.
Ran vCenter converter on the win7 machine, saved to external drive, booted VM, ran BCUninstaller to remove all the extra gook, cleared all the unused drivers out, nuked other profiles, and did tidying up of temp/cache/blah. Saved ~ 200Gb of space.
Now it's spool up the win7 VM and run Unity mode to do work, while I curse and install all the updated new vendor stuff on win10...
Weirdest setup I've done with archaic stuff was the DOS networking drivers, an ISA NE2000 card, and I think an IBM(?) networking utility to do virtual disk drives for retrogaming.
In the early 80's I was working as a technical consultant for a mainframe manufacturer. Diagnostic visits to customers were often out of hours and a 2 hour session starting at 6 was often deferred until the batch work had finished at midnight and then took many more hours of diagnostic work to identify and then resolve the problem.
If the overnight stay were unanticipated we were allowed to book a hotel room outside the expenses guidelines and were encouraged to pop down to the local M&S to buy underwear, socks and a clean shirt before attending the customer site.
It made an unplanned overnight stop a bit less hellish. Most of us did take a bag on the off-chance it would require an overnight stay and where the issue was critical and required several skill sets a late evening meal and a few beers on the company with fellow engineers always made the pill a bit sweeter.
I supported an ECRM VR30 Imagesetter connected to a Harlequin RIP running on a Windows 3.1 box. on a rather old-school, well kind of state of the art system in 1990. Box its self was a 486 DX50 EISA motherboard, 640K RAM, 512 MB HD, a proprietary SCSI card, and a 3COM 3c509 network card. That network card was a pain to setup too. That required editing the net-bindings in the net.cfg file in order for the COPSTALK driver to find it.
Windows 3.1 sat on top of DOS with Microsoft networking drivers, along with COPS-TALK - aka Cooperative Printing Solutions Mac-AppleTalk emulator, which was needed to print to the RIP since the box had to emulate a Linotronic 330 imagesetter and that only had an AppleTalk interface. COPS also gave the few Mac users file sharing capabilities with the PC's in the office too so it was a win-win in that regard.
With 640K of RAM, there was about 50K left after loading all the drivers, and I wondered if the RIP software would run at all with such a tiny amount of RAM. I'm not exaggerating here, it was only 50K left! I inquired why I couldn't load any drivers high, and use WFW to replace the Windows 3.1 sitting on DOS with Microsoft Networking. The ECRM tech said that it was due to the drivers not working when loaded high, and that WFW was too busy chatting on the network, which caused buffer under runs on the imagesetter due to the SCSI interface timing out. A buffer under run is something we didn't want because the film cost $250 a roll, and the boss would not be pleased to hear that we were wasting film.
Once loaded, however, we said our prayers to the Oracle and left chicken entrails on the ground in front of it to ensure it would run, and it did. The setup was pretty stable and ran like a champ for the next 20 years. We finally retired the VR30 and RIP when film and chemistry got too expensive and the chemistry spoiled before we could use it all. At that point, we switched to a Postscript L3 Xante printer.
LOADHIGH and Dr Watson, haven't heard that for many years.
I do remember upgrading PC's from 640K to 1M by putting extra chips in so we could fit the NE2000 and load the Netware client.
Shame about Netware, it was super stable, and now a cheeky reboot has somehow become acceptable.
Still neck deep in infrastructure and loving it :-)
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
