How a single buck bought bragging rights in the battle to port Windows 95 to NT A former Microsoft engineer has waxed lyrical about how he and a colleague made a sporting bet over how far a new build of Windows would get before crashing. Clue? Further than you might think. Plummer, the programmer that worked on MS-DOS 6.2 and Windows NT, yesterday explained how a dollar bill came to be affixed to a …
Both operating systems were very easy to use originally, eventually I saw some issue with NT running an independent mail server but initially every function issue was very easy to fix - the tech developers at Microsoft were very efficient back then, they are not worse these days they are just working in a much worse software and malware environment. Eventually we started to see a few DLL hacking efforts but running everything behind a firewall box (not thinking that firewall software might always work) stopped all the attacks - but attacks were not common back in the old days.
I've still got the original multi user NT installation, I might be able to install it again to play with it but I need to find a PC with both a floppy disk drive and a CD drive.
I know, rose-tinted glasses and all, but Windoze was ALWAYS garbage. Back in the day, you didn't need malware, it would fall over all on its own. 95 would crash with a sideways glance, NT took a tiny bit more, but that crapos was unstable.
And of course back in the day there were REAL viruses for Windoze, no user interaction needed for them to replicate.
You wanted stability on the desktop back then, you used a Macintosh. On a server, you used UNIX or Linux. You wanted utter shite, you used Windoze.
I'm sure there were plenty of cases that would lead to kernel panics, or just ways to crash important daemons; but IME even back then Linux was more stable than any existing Windows version, and for some common situations (SIQ!) more stable than OS/2. (Though you could get around the SIQ issue if you had remote access to the OS/2 system and could kill the offending process...)
Of course, whatever applications you were running on any of those systems might be an entirely different story, and all of it was riddled with security vulnerabilities. Levy's "Smashing the Stack for Fun and Profit" only came out in 1996, and prior to that, and similar attempts to publicize exploit development, there was still widespread belief in the industry that people like Morris were rare geniuses and exploit development was so difficult that vulnerabilities could mostly be ignored (despite all the evidence to the contrary).
Yes, there were plenty of ways to crash the thing, but if you were using it for work, why would you do that? Kinda like intentionally hitting your thumb with a hammer, no?
Security on those early systems wasn't as much of an issue as it is today with near ubiquitous always-on Internet connections. Back then, for Internet access folks were still using dial-up into shell accounts if they were lucky, or dial-up UUCP if they weren't, and many were still stuck in BBS land. (AOL was still Q-Link, and barely existed at the time). They also had more clues than today's users ... they knew how the system worked. They had to, just to get online with it ...
"smashing the stack" was a compilation of known vulnerabilities and techniques, somewhat similar to Schulman's "Undocumented DOS" series, which was far from being undocumented[0], and was in fact quite well known. I gave an occasional talk/lecture on the subject (using buffer overflows etc., not DOS system calls) at both Berkeley and Stanford in the late '70s and early '80s. Here's a link to the Phrack paper:
https://insecure.org/stf/smashstack.html
tl;dr version: If you put 10 pounds of sugar into a 5 pound bag, you'll be able to stand on the resulting pile and reach the cookies on top of the fridge.
[0] Well, undocumented by Microsoft, perhaps ... but on Usenet and various BBSes, maybe not so much.
I used to download Linux from the University of Helsinki FTP site onto a tawdry* of floppies, circa 1993. No desktop, but stable.
Recollections of Schrodingers Windows at the time were less stable. Simply by observing the desktop something would change in the kernel and it would turn to soup.
* That may, or may not, be the collective noun for floppy disks.
I tried slackware back then. Gave up on it because out of my PC, a pretty generic box, this wasn't supported and that wasn't supported and I never did get as far as a desktop as it would crash spectacularly trying to bring up the graphical display (yet some game or other, I forget what it was, had no problem).
Unfortunately the following year they released Windows95, and several service releases later it was still a complete piece of crap. Never before has an OS had the audacity to fail completely (blue screen) at something as mundane as the user ejecting a CD ROM.
It got better though, 98SE worked quite well (as well as the Win32 ever would), and I think it reached peak Windows with XPSP3 (whose real problem was the dumb decision to make all users administrators by default, which took any sort of security that it may have had and tossed it in the bin).
Windows isn't great so much as simple to use and widely supported. Simple to use means you can get most tasks done with pointy clicky, no need to drop to the command line to issue arcane incantations. And widely supported means stuff mostly works (though bad drivers and drivers buried in tonnes of bloatware do exist, take a look at most printer drivers).
As for dominance, that depends upon how you define it. If you're looking at "computing devices that people use" then it's probably Android. If you're looking at "computing devices that people own" then it's probably Linux (quietly lurking away inside smarter printers, security cameras, routers, etc etc etc etc). It may be Windows on the desktop, but these days it has strong competition from both Mac and Linux, as operating systems become more generic so rather than using X because Y, there will be other considerations such as online security and "does it work with my tablet?".
Suffice to say, when you die, Windows probably won't be a dominant force any more. I think an increasing number of people (myself among them) have more or less given up on big power hungry desktop machines for mobile devices and, well, Windows just doesn't feature there at all. Microsoft poured huge amounts into mobile Windows and walked away with its tail between its legs.
Yes, for my needs it was stable. Far more so than the Windows of the day[0]. I would sometimes go a couple weeks without a reboot on my primary box (and then it was usually only because I was watching and sometimes contributing to kernel development and felt an update was in order), and I had a reclaimed-from-a-dumpster 386SX print and file server that stayed up for well over 6 months at one point. I don't remember more than one crash where I lost work ... and that was my own damn fault[1].
Why Linux? Because it was there. I came from the BSD world, and was still heavily involved in it. For a home un*x I had first run a battered old PC 7300 (so-called AT&T UNIX PC) hacked into a 3B1. That was followed by (and concurrent with) Coherent and Minix, until Net/2 was viable, then 386/BSD. Throughout, I also had dial-up to more robust systems, but even local calls still cost money during the early part of that timeframe. I had fiddled about with very early Linux (it was not ready for prime time, by any stretch), but for some reason the first Slackware release caught my eye in August of '93. I downloaded it, installed it on the above dumpster-derived-386 (with 8 megs, math-co, VGA and a 80meg HDD! WOW!), and dropped both Pat and Linus an email. The rest, as they say, is history.
[0] I used DESQview when I needed that kind of thing on DOS ... Windows was a joke in comparison.
[1] Hint: Never run untested development code on your primary desktop machine when you have a day or so of un-saved contract open in your editor!
Correct! WindowsNT would run for months without a reboot on good well supported hardware. Remember these OS's were used on all kinds of hardware and Microsoft couldn't write drivers for every one of them. If you used shitty hardware with shitty drivers expect your OS to crash on the slightest whim. Windows95 of course was a different story - a dodgy CD in the CDROM drive or a bad sector on your floppy was enough to stall the 'OS'.
Linux only fared better as all the drivers for Linux were built into it and not written by crappy 3rd party developers. If the hardware wasn't supported - tough luck.
"They are all just Operating Systems, not religions or philosophies."
However the philosophy behind the operating system makes a difference to me.
If one OS exists primarily to make a profit for the shareholders, with the needs of the end users deemed unimportant, and another OS exists primarily because those end users saw a need, and so provided it for themselves and made it freely available to anybody else who had similar needs .... Well, I know why I run Slackware and BSD instead of Windows or macOS (or RedHat or SUSE or Ubuntu ... or even Debian, where internal politics vastly outweigh the needs of the end users ... ).
But I'm a testimonial, a sample of one. Use whatever you like.
That was the year it was already being developed for its second architecture after x86—the DEC Alpha.
So Linux went 64-bit the same time it went portable. Windows didn’t go 64-bit until years later: even its Alpha version ran strictly in 32-bit “TASO” mode.
Linux currently runs on about two dozen major processor architectures, which is not far short of one for every year it has been in existence. Meanwhile, Windows is struggling to manage a second architecture that is not x86.
Back then the Windows nt 3.51 was the best operating system EVER and we were runnind thousands of users off it by the time the "Chicago" user interface came to it in tne form of Windows 2000 which we were already running in early Beta by 1997.
We
I may be off by a year, but im pretty sure i was running a linux desktop ( on a 486 ) in 95.
In 97 id adopted a laptop that the sales team no longer wanted. P133, huge ( for those days ) 256 color screen on which i was running linux. Slack i think. Installed from a stack of floppies.
Windows and Linux were both installed from stacks of floppies. I used to shut my machine off and move the IDE cable from my Windows hard drive to my Linux hard drive. Then I guess I finally only had to reboot using lilo, which was very nice, and choose which OS. In about 1997 I was installing various OSs on Cartridge drives and would boot up whichever I was testing by changing the cartridges. Not the zip drives, but the bigger ones.
> I was installing various OSs on Cartridge drives and would boot up whichever I was testing by changing the cartridges. Not the zip drives, but the bigger ones.
Jaz drives? I had forgotten them until I read your comment. As I recall they were quite expensive for what they were, at least compared to CD-Rs (and a little later, -RWs).
You wanted stability on the desktop back then, you used a Macintosh
1995 was the era of Classic MacOS 5.
Classic MacOS used co-operative mutlitasking with no memory protection. Just like Windows 3/95/98. And if you used Classic MacOS, you'll be all too familar with the Bomb of death (rather than the Blue Screen Of Death)
There are two reasons I never liked Apple, both of which were valid from the first Apple machine to now. 1) There is nothing intuitive about the Mac, or for that matter, anything Apple. 2) I could never afford it. I go for power and bang for the buck. Pretty is nice, but I need bang for the buck.
I'm no Apple fanboi but I do have to use one for work. I find Apple exactly as intuitive as Windows once you're used to them, both take time to get used to how they do things.
On other things, MacOS has a faster filesystem than NTFS in my experience but a tendancy for things to "just not work" like missing keyboard control or external hardware not made by Apple just stopping working where it would be fine on Windows.
It's still not intuitive - try it with multiple monitors and it is all over the shop.
Full disclosure - I have a Mac Studio. Windows just seems to be much better at the productive stuff, not the cool stuff. You can switch between 2 windows cutting and pasting from a Word Document/website into a spreadsheet so fast in Windows while a Mac ponders over it every time you switch apps.
NO GUI is intuitive. The meanings of the icons and user actions have to be LEARNED. Fortunately for us, MOST GUIs work sufficiently-alike that after you've learned one of them, you can easily learn another.
(Yes, you in the back waving your hand -- I see you, and your copy of Plan 9 running on your Toshiba Tecra with its 800x600 screen, PCMCIA slot, and single USB 1.1 port. That's why I wrote "MOST GUIs".)
> You wanted stability on the desktop back then, you used a Macintosh
Err no, for serious work, it was a Unix workstation: Sun, SGI etc. namely the platforms NT was pitched to compete against.
Distributions like Sun’s Solaris for x86 were attempts at putting a serious alternative to SCO Unix desktop on the x86.
Yep. Macintosh systems were good because they could load up large image files and passably work on them without grinding to a shuddering stop and then crashing which is what happened whenever similar was attempted in Windows. For some, utterly explicable reason, image editors didn't want to spend their life losing work and restarting their computer just because they had the audacity to try to edit a large image file... therefore they used Macintosh systems. The alternative was extremely overpriced Sun, SGI, etc workstations running bespoke hardware and software and the suppliers of these systems were more interested in maintaining their cash cow lock in than actually improving, and opening up, much.
Except that Windows was made for business purposes, obviously, since it does have the IBM heritage. Macs were, in fact, better for running Photoshop back then. 3 people made money doing that. All the others I knew lost $10,000 buying stuff that they never got paid off. And now, there's nothing a mac or linux box does for an average user than a PC doesn't do better for still way way less money. And no, you aren't going to install Linux on your non-technical household type users. They just want to play a game on a PC. I don't need unix or a mac to run The Sims.
And now, there's nothing a mac or linux box does for an average user than a PC doesn't do better for still way way less money.
1. Linux runs on the same hardware as does MS-Windows and MacOS.
2. Linux is available free-of-charge; you have to pay for MS-Windows and MacOS (though it's included in the price of new Apple hardware, and in the price of pre-loaded commercially-built retail x86 systems).
So, on what basis are you claiming, by inference, that Windows systems cost "way less money"?!!
You can always spot those whose knowledge comes from reading a post by someone whose knowledge comes from reading a post whose knowledge comes from Microsoft marketing* - or maybe a bit less direct than that. One thing they don't realise is that those of us who run Linux as a daily driver are often asked to sort out problems for those running Windows.
* The alternative is those who, despite never actually having installed Windows, have used it, now consider themselves a computer expert and made a ham-fisted attempt to install Linux by clicking all the non-default, wrong options.
For a business $100 for a Windows license is nothing compared to how much they would have to pay for a full time IT support guy to manage their Linux systems - believe me, I've spent weeks trying to configure Linux servers myself rather than hiring someone to do it for me. I did learn a lot, but businesses aren't doing it so their employees can learn these skills on their time.
"And now, there's nothing a mac or linux box does for an average user than a PC doesn't do better for still way way less money."
Yup, when it comes to crashing, slow boots, extremely slow and buggy upgrades a PC with Windows is way ahead of the the pack, or at least, way ahead of Linux. I don't have a Mac so I don't know for sure but I'd guess on those metrics Windows is ahead those as well.
"And no, you aren't going to install Linux on your non-technical household type users."
Odd you should say that. I installed Linux on my cousin-in-law's PC some years ago after she'd got hit with ranomware. She's still using it for web, email, maybe a bit of LireOffice, maybe not much these days, very fond of Google Earth. She's at least 90 and a in terms of long-ago work experience, used to be a hairdresser. Does that fall into your non-technical household type?
When the Windows 95 "Chicago" user interface finally came to WIndows NT 3.51 which was what is now known as Windows 2000, we were able to finally get WIndows up on multiple systems such as PowerPC, DEC Alpha and x86 Intel/AMD architectures. Were were running Windows 2000 early betas by 1997 so we had first-hand experience as to what it could do multi-platform-wise and it was WONDERFUL! The DEC Alphas were FAAAAAST Running 220 Mhz at the time AND we actually got a 64-bit version of WIndows 2000 running on it so we could address huuuuuuge amounts of RAM and hard drive space. YES! There was 64-bit version of Windows 2000 and it wasn't just for the Intel Itanium chip! There was a 64-bit DEC Alpha version of it too!
In fact, we STILL run Windows 2000 on a lot of our x86 32-bit server motherboards as near-space and deep-space monitoring systems! Some of those machines have NOT been turned off since 2002! (i.e. 22 YEARS continuous!) -- It also shows the POWER of those old Gigabyte and Tyan Motherboards where the Capacitors have not yet dried out and the RAM chips are still good! Even the ridiculously expensive 120 Gigabyte SCSI hard drives ($3000 USD per drive!) we had back then are STILL running! (there was a 128 Gigabyte addressing limit on hard drive space back then!). We use them for our custom-coded SDTV deep-space-imaging software using our STILL RUNNING DPS Perception video cards at 30 megabytes per second uncompressed 720 x 486p SDTV video out past Geo-SYNC orbit which compresses the original video to a custom interframe-compressed-wavelet file and to a 2D vector file for eventual Earth basestation download. (we modded the DPS Perception cards to do non-interlaced 30 fps 720x486p video!) We rad-hardened everything by stuffing all electronics into a big 3/4 inch thick Tungsten Faraday Cage box!
We aren't scheduled to replace them to 8K/16K video until about the year 2035 (i.e. after 33 Years of continuous operation once the Satellite solar panels finally degrade past their lifetime!) -- We have much better video systems in 2024 and a proper SSTO to do the actual deployment BUT we haven't bothered as we still get our 6 times daily downloads (every 4 hours!) with no hitches and we can expand the 480p to 4k resolution using a fractal-based video frame resizer program. P.S. We also see a veritable L.A. FREEWAY of "Fast Walkers" up-there at 440 miles (past ISS) to Geo-SYNC orbit in case you're wondering!
Sooooooo, Windows has worked VERY WELL for many users! We are STILL running Windows 2000 on some machines after 22 YEARS which shows just how good it really was and STILL IS!
Not many Linux Servers can say the same thing!
V
I remember the DPS Perception Video Recorder cards. They were awesome in their day. The fact that they recorded directly to a SCSI drive, bypassing the system bus was genius. It made desktop video editing possible before PCs were really powerful enough.
I remember the driver would do all kinds of cool stuff. You could access the video storage drive as another drive letter. It would also make directories available with each video frame as a separate image file (Targa, if I remember correctly).
We used Newtek Lightwave to render 3D animation directly to the PVR drive, then we could print it to 1" Type-C video tape through the analog video outputs.
Very expensive stuff back in the day. Now, a budget smart phone blows it away!
In fact, we STILL run Windows 2000 on a lot of our x86 32-bit server motherboards as near-space and deep-space monitoring systems! ... Not many Linux Servers can say the same thing!
I'm pretty sure there are a hell of a lot more Linux servers (still) running, here on planet Earth, than there are Windows 2000 systems in outer space. There probably are even a hell of a lot more Novell* servers still running here on planet Earth, than are Windows 2000 systems in outer space.
*We could never convince users whose PCs were running NPRINTER to not power their PCs off at the end of the day. No amount of signage or training would fix this human-based computer problem. So, we (IT) decided, "fuck it", and declared a new policy: if a department wanted a printer added, they would fork out the cash for an HP JetDirect. Cry and plead penury as you will, we will not "go cheap" for you and install an NPRINTER-based printer for you, because we (IT) will pay far-and-above the money you saved in hardware costs, in tech salaries troubleshooting multiple "I can't print" problems due to a switched-off PC workstation/print server.
"In fact, we STILL run Windows 2000 on a lot of our x86 32-bit server motherboards as near-space and deep-space monitoring systems! Some of those machines have NOT been turned off since 2002! (i.e. 22 YEARS continuous!)"
Any public information about this installation available on the Web or anywhere?
I have an interest in 'durational computing' and long running systems.
PS: As a humble end user, the Win2k computers were a revelation when they arrived at work compared to the ME ones. Just bog standard desktop boxes.
You need to realise that there's and argument consider W2K was probably peak Windows and everything has been going downhill since then but you might have trouble running more recent versions on your old H/W anyway. It's always good to hear of H/W that was built to last, so much is built to be disposable these days. If you wanted that reliability of S/W there day's you'd be well advised to go for Linux or BSD,
In those days, for a server you used Novell Netware. Servers ran for years without reboots. You could even patch it without rebooting.
Some server hardware at the time would allow hardware replacement without rebooting. I once replaced a network card without rebooting the server. You go to the command line, unload the network driver, pull the NIC and replace it, then go back to the command line, and load the driver for the new card.
I once was called in to work on a server that was showing an uptime of over 6 years!
"In those days, for a server you used Novell Netware."
Oh no I wouldn't. I'd use a real Unix server. I had a very brief encounter with a Netware server. Trying to stop a database server brought down the whole thing because, as far as I could see they were doing the equivalent of running the kernel and all services as a single process.
Win XP almost worked. At least I could kill a meandering app without a restart. But by then my Linux box was still wild but tamable and had some apps, so I switched (old journalist/engineer) and I "bent it to my will". Never looked back. About 2002..."These days, kids don't know how lucky they are..." or something.
I am a self-confessed Apple fanboy and have been using Macs, along with other Windows and other platforms, my entire life, and I promise you that in the 90s if you wanted stability on the desktop you did not pick a Macintosh unless your workflow never involved hitting the "File" menu, hovering over "Print" and saying a prayer before letting go.
Windows NT 4 was a curse made into an operating system.
The sheer ridiculous incompetence in how many times basic settings could no longer be changed after the installation of necessary service packs was staggering. Want to change a particular network setting? (Too long ago to remember the specifics, and I've blotted this kind of stupidity out of my mind). First, uninstall the latest service pack. Then make the network setting change. Then reinstall the bloody service pack. The sheer unbridled incompetence of Microsoft in vomiting out such tripe is something only matched by their continuance in still doing this kind of thing decades later.
New, USB-attached external CD/DVD ROM reader/writers are available for about £25~30 + VAT. And I picked up a used, Dell USB-attached external floppy-disc drive the other day for £12. (Now I just need to find some new old stock floppy discs ...)
> might regard peak NT as Windows NT 3.51
Indeed. Though perhaps it was due to it being the only Windows release with a feature freeze for ~9 months where they focused entirely on bug fixes. This opportunity only came about because they were waiting on IBM to release their PowerPC hardware.
Some more info here:
https://web.archive.org/web/20100604082534/http:/www.winsupersite.com/reviews/winserver2k3_gold1.asp
I had a Explorer 95 style "preview shell" on NT 3.51. The eventual NT 4.0 was less stable and easily BSOD by a dodgy printer or graphics driver because of the stupid decision to put GDI into the Kernel for about 10% performance increase on video in an era when performance was doubling for video in less than a year. Stupid. So till late 1990s we continued to use NT 3.51 for servers.
Even the 3.5 to 3.51 was due to win95 adding Win32 calls so Office 95 wouldn't work on Win3.11 / WFWG3.11 with Win32s installed, because Office 95 wouldn't work on NT 3.5.
It took IT and MS years to recover from Win9x which should never have been sold to businesses. It was best for games.
It took IT and MS years to recover from Win9x which should never have been sold to businesses
A lot of software didn't work on NT4 when it first came out - only Windows 95. Once we could get drivers & software for NT4, it became obvious it was a way more reliable OS than Win95 and we started to move to it. We still stumbled upon the odd piece of software which would only work on Win 95/98, but NT was the future for business.
the stupid decision to put GDI into the Kernel for about 10% performance increase on video in an era when performance was doubling for video in less than a year. Stupid.
1. MS needed to sell NT to a market which wanted to keep using the hardware they already had. The "Go buy a new video card" attitude might work for home users who want to run the latest killer FPS games, but would not work with a business with tens, hundreds, or potentially, thousands of computers (our network had about 11K PCs at the time).
2. We can look back, now, and see that video performance was doubling-or-so in less than a year. It was not obvious, back then.
Oh, I don't know.
I was quite fond of both Vista* and Windows 7. It seemed to be time when MS were actually trying to make an OS that worked, was reliable for users by not BSOD'ing if you as much as glanced at it the wrong way and looked quite pleasant on the eye. It feels like it's been downhill ever since.
I think we look fondly on 2000 because it was the first business class OS from MS that had things like plug and pray.
*As long as you ran it on a machine capable of it, of course. All the "Vista Ready" bullshit was...well just that. Bullshit.
This will certainly get me both bashed and downvoted, but with Windows I'm most fond of the version that allows me to get the job done.
For me, these last few years, that's been Windows 10. Yes, I know you guys hate it. But it is far more stable than Win7 and offers less compatibility issues because they've been worked out since Win7's time (remember how many times you had to use Compatibility mode?). I'm sorry, but I will stand up and say Win10 is the best so far because as a user you can pretty much just get on with it and do your job: hardware drivers are there, software compatibility is almost a given, stability is good. I don't understand the fascination with the Program Manager - ugh. Nice to visit in RetroLand, I don't want to live there for years.
Icon: mine's the asbestos version, thanksmuch.
And 7 is a re-wrapped Vista. That's how it's generally gone with Windows releases. They have major releases like every 3-years (Win95, Win98, XP, Vista, 8, 10) and then minor releases (98SE, Me, 7, 8.1, 11) about a year after the major release, which were always just warmed over versions of the major release. It's been a bit different since they moved to a more SaaS model with Win 10 and 11, but you can still point to things like the Windows 10 "Creators Edition" as being a significant update compared to some of those before it.
I see Jellied Eel has clocked in for the day and is going straight to cyberstalking me again. Remember everyone, no matter how bad you think life is, it could be a lot worse. Just look at Jellied Eel and realize how much further you could fall. A like third or fourth string vatnick troll who has to work the graveyard shift. Not even good enough to be assigned to any of the bigger name sites, they're relegated to trolling niche sites like El Reg.
Just because they aren't trolling doesn't mean they aren't here. They seem to have taken to stalking me across every comment section on El Reg and downvoting every single comment I make. A couple of times I've gone to check the comments because that's the fastest way to check if there are any replies, and will see multiple comments across multiple stories all get downvoted at more or less the same time. So, if it's not the vatnick eel, it's someone equally pathetic trying to pull a Reddit-style karma bombing.
The projection seems strong here. I'm not advocating for shit, but the fact that you think that's what I'm doing means that you're exactly the sort of person I'm talking to/about.
I'm tired of evangelists of every stripe. I don't care if you're some god-bothering moron, you think electronics manufacturer X is better than electronics manufacturer Y, or you can't shut up about some specific bit of computer software. At the end of the day, they're all born from insecurity on the part of the evangelist. Face it, if you were secure in your decision that would be enough. You wouldn't need to try to "convert" others to your belief system. The very effort means that you are relying on a bandwagon-style logical fallacy to derive your self-worth. The bigger a particular community, the more likely it must be to be The One True WayTM. Just get a fucking therapist like a normal person. They can help you work through these issues so you don't need to bother everyone else with them.
If you use and like Linux, fine, I really don't give two fucks. It makes zero difference in my life what OS you use. If you're happy with macOS or Windows, good for you, still don't care. Maybe you've abandoned desktop operating systems completely and have gone fully mobile. If that works for you, great, I just don't need to hear about it with every fucking post you make. I. Don't. Care. I don't run around proclaiming "Windows is the greatest OS evar!" because it's not. It's the best choice for me, given my particular needs at this point in time, the end. I don't get emotionally attached to a specific piece of software. It's a tool to do a job, nothing more. At points in the past it was Mac OS X or Linux, and so I used those until things changed. It may change again in the future and I move back to Linux or macOS, who knows. I've had Android phones, I've had Windows Mobile phones, I've had Blackberry phones, and I've had iPhones. I evaluate what's out there against what I am looking for in a device, and what I anticipate wanting out of a device for the next couple of years, and choose accordingly.
Better to be a reed that bends in the wind than a tree that tries to stand firm and is blown over.
Nah, 7 was probably their best release ever. Better than Worsta by far, and better than the spyware-infested 10 and 11. The 8 disaster should never be spoken of again.
But hey, it's been fun to watch from the outside. Windoze is useful for games, and pretty much nothing else. It's the thing that you install because some fool used it instead of Linux for their industrial control system. It's the thing that you rejoice forcibly removing from your network when either a Mac or Linux replacement option appears.
I've seen a machine that seemed like it was on its last legs with Worsta have 7 installed, and suddenly it's functional again. (And then I reformat and put Linux on it, and it's like a new machine.)
And, like always, just another bunch of repeating what the "cool kids" say and not a single original thought to be seen. Just another poser who thinks simply installing Linux means they're some l33t hax0r, when they go out of their way to find a distribution as close to Windows as possible, then piss and moan endlessly about everything that's different from Windows. Be honest, you wouldn't be able to tell the difference between BASH or CSH, would you? You probably don't even know what those are.
Grow up nd maybe spend a little time learning what it is you don't know, because it is legion.
I only have but one upvote to give, so you'll have to make due with a thumbs up icon as a bonus for a refreshingly pragmatic comment. Whatever works best for you is, pretty much by definition, the best version of Windows for you. It's not healthy to get emotionally attached to any specific version, or to inanimate objects period. Exceptions might be made for things given to you by a deceased relative or friend, which are one of a limited number of mementos you have from them, but getting your panties in a wad over one version of an operating system vs another is really just the height of stupidity.
> Windows 10. ...is far more stable than Win7 and offers less compatibility issues
My Win7 is as stable as a Corgi in the kitchen. About once a year (it runs 24/7 and works 10/day) it hangs-up for a minute and then blames a video card error. I used to could send a report, futilely, but now it won't go (MS probably closed the report box). Yes, I have updated the Intel chipset drivers. The win10 machine a better brand, has similar trouble in much less run-time, +++ is far more aggressive about low-quality "updates". Seems like if I didn't update it yesterday, today will have to wait while it plays with itself.
"that allows me to get the job done."
Yep. And that was always Win2K for me ...until January 1, 2010, when I stopped supporting all things Redmond.
Today, I have one Windows box. It's Win2k. It has one job, the same job it has had since early 2000 ... it runs AutoCAD2k. The box hasn't given me any trouble at all ever since I got the plotter drivers sorted out in March of that year, even though I use it at least weekly (I use her to document everything I've built or fixed/upgraded on my various properties ... I honestly never saw a reason to change.)
Yes, she's airgapped. Duh. So is the BSD-based file server she's attached to.
Sorry 7 gave me WAY less headaches than 10, 10 & 11 still periodically exhibit some otherwise unreplicatable error that sometimes even a total reinstall won't fix, needing resort to web searches and all manner of dism commands and sacrifices to the gods of electronica in the hopes it will decide to work.
Program manager I agree on though, I don't remember windows 3.11 with any fondness, it being less advanced than the Amiga OS...
Vista is a very unfairly maligned OS. It made a lot of really significant and important improvements, but most of them weren't user facing, so are generally ignored/forgotten about.
It had
1) A completely rewritten GUI that made use of 3D hardware acceleration in graphics cards, unlike GDI+ which was all software rendered
2) A process scheduler that understood the difference between multi-core CPUs and multi-CPUs, which is a subtle, but very important, distinction
3) Significant improvements in security, particularly in hardening of the driver model (of course all people remember is that they couldn't use their old 2K/XP drivers, and some hardware vendors were slow to release Vista drivers, and everyone forgets how for like 2-3 years, literally every week XP had some new critical vulnerability found [sometimes more than one] that required immediate patching)
I kind of wish MS had stuck with their original plan to rewrite the bulk of the OS in .NET, and some of the other things they jettisoned along the way, but Vista was still a very significant release and is very unfairly maligned by people. Generally speaking, if not for the major changes made in all the versions of Windows everyone claims to hate, the versions they claim to love never would have been possible, which includes Windows 7, a barely warmed over Vista that just aped Mac OS X's Dock and toned down the glass effect of the Aero UI. Win 7 had a few other changes, but they were just minor improvements on what Vista started.
You are quire correct, however, that the Vista Capable or whatever they called it, was a self-inflicted blunder. No doubt it was because OEMs like Dell wanted to keep shifting low end PCs with shitty IGPs that weren't up to the task of Aero, and Microsoft capitulated. If you had a more mid to high end system, Vista was just fine. Windows 8 was sort of the same deal. If you take away people bitching about the shoehorning of a touch-oriented UI onto a desktop version of the OS, there was precious little else that people complained about with the OS. Shit, Microsoft even did what people always claim they want them to do, and reduced the size of Windows by a couple GB during the Win8 years, not that you'll ever hear people give them credit for it.
There's a whole groupthink psychology to Windows releases. A lot of people claim to hate Windows <Whatever> simply because other people say they hate it. They themselves probably never even used it, or have no particular complaints about it personally, but they want to fit in with the group, so mirror their speech. And if you take the time to actually look at the complaints people raise about Windows <Whatever>, 99.9% of the time, it's because Microsoft changed something visually. They moved this or that setting to a new location, or a button moved 1px to the left. Rarely do you find people complaining about actual technical issues. And the few times you do find someone saying how the OS and/or apps crash a lot, when you press them a little further, they've never done anything like run memtest or check for malware. They just leap straight to blaming the OS. There are plenty of technical deficiencies in Windows, just like Linux, macOS, *BSD, and everything else out there. Microsoft and Apple have both slashed their QA staff and things have gotten progressively worse since, and QA tasks are rarely the sort of thing that interest FOSS devs. But, even on sites like this, which cater to more technically inclined users, rarely do you see anyone who has taken even basic steps towards debugging the issue to show that it really is the OS.
You're mostly correct, but there's a few other factors.
Vista, regardless of its benefits, was rather buggy until SP1 - but by that stage it was too late.
XP was out for five years prior to Vista, the longest time between releases, at least until the gap between 10 and 11 (which probably doesn't count, given the number of revisions of 10). People got used to a mature infrastructure, despite its disadvantages.
Aero was an advantage but it's an issue when people can't use their old drivers, and it took time for driver quality to equal that of XP, in the same way XP/2K drivers initially were variable in quality. Also, IIRC, the WDDM version in Vista kept two copies of the display in video memory at the same time, drastically increasing VRAM usage. This was an issue with many cards available at the time, until 7 kept only one frame buffer in VRAM.
The limitations of the IGP were one issue, but that could be solved by disabling Aero. Unfortunately many 'Vista Capable' systems also had inadequate RAM, and that really impacted on performance.
The improvement in security was not welcomed, especially by many application writers who had to be dragged kicking and screaming into the new world where they couldn't run as administrator all the time.
Multi language support was *much* better in Vista than prior releases. I was daft enough to buy the Ultimate edition of Vista for the multi language support.
Once it reached SP2 Vista was rather nice. It supported 64 bits properly, the driver situation wasn't as moribund as running XP x64, it still featured components such as the Removable Storage Manager, and security could be configured to be improved even over defaults by switching to the secure desktop when displaying prompts. Gadgets were quite cool too.
There haven't really been any poor NT based releases. 8 comes the closest with its substandard mobile first interface that was never fixed, but technically was an improvement over 7 underneath. 11 is certainly usable, albeit with niggles, too much movement away from user centric computing, and unnecessary default hardware requirements, but it's stable and usable.
The 'worst' NT release is quite possibly NT 4, the first Windows OS I used as my primary system after migrating from OS/2. A lot of that is due to rapid technology changes between 1996 and 2000 and a slow release cycle; by the time 2000 was released the limited Direct3D support in NT4 and the complete absence of USB support were becoming real hindrances to using consumer focused applications or games.
Come to think of it, isn't NT4 the one that has like a recursive loop with the service packs? You can't install one of them unless you have a certain version of IE installed, but somewhere along the line that version of IE stopped being downloadable and the only version you could download required the service pack you were trying to install. And this was before SPs were cumulative, so you couldn't just grab whatever the latest one was (didn't they get up to SP7 with NT4?) and install that.
Yep, that was pretty much the level of incompetence foisted on NT4 administrators by Microsoft. Want to change a particular network setting? No problem, just uninstall the latest service pack, make the change, then re-apply the latest service pack. Add in further moronic dependencies such as IE and the desire to book a flight to Redmond and impart, directly, using a clue-by-four, the frustration inflicted by the incompetent developers and QA teams at Microsoft.
"You are quire correct, however, that the Vista Capable or whatever they called it, was a self-inflicted blunder. No doubt it was because OEMs like Dell wanted to keep shifting low end PCs with shitty IGPs that weren't up to the task of Aero, and Microsoft capitulated. If you had a more mid to high end system, Vista was just fine. Windows 8 was sort of the same deal. If you take away people bitching about the shoehorning of a touch-oriented UI onto a desktop version of the OS, there was precious little else that people complained about with the OS. "
Agree with you on Vista (although the over-obtrusive elevation requests were also irritating, and 7 improved on this). 8 was a different situation though - the UI was absolutely awful: that home screen with the big buttons and the desktop relegated to an adjunct, the idea of separate full-screen and desktop programs (including some, such as IE, where both existed), plus the general ugliness of the UI. From a usabilility perspective it was a disaster.
While I do agree the prompts about "do you really want to do this" were annoying, when you take a second to think about it, they were kind of a necessary evil. They were intended to try to force people to stop and think, "do I really want to do this?" This was coming off the golden age of spyware/malware, and people would just install shit with admin privileges without so much as a second thought. Vista was a little bit too much, Win 7 was maybe a little bit too little, but by 8 they had it pretty well dialed in. Sort of like how it took them a few tries with Windows 10 to deal with the new policy of forcing updates that may require reboots. Yes, it's annoying since it seems like it always happens right when you're in the middle of something, but if you stop and consider, a lot of the shit that we used to deal with (email worms, the messenger service reboots/spam) have almost completely stopped. Grand scheme, it seems like a small price to pay.
"Sort of like how it took them a few tries with Windows 10 to deal with the new policy of forcing updates that may require reboots. Yes, it's annoying since it seems like it always happens right when you're in the middle of something,"
I remember that one. At a local archaeology do, somebody was about to demonstrate his surveying S/W - luckily it was just to a few of us after his talk. His laptop must have picked up an open WiFi or something and phoned home to be told to update there and then.
"but if you stop and consider, a lot of the shit that we used to deal with (email worms, the messenger service reboots/spam) have almost completely stopped. Grand scheme, it seems like a small price to pay."
It only seems a small price if you don't know there can be a smaller price to pay by doing unobtrusive background updates that only occasionally go low enough to require a reboot after the upgrade's complete, and just plain reboots at that, When you know that it seems an inexcusably inordinate price.
It only seems a small price if you don't know there can be a smaller price to pay by doing unobtrusive background updates that only occasionally go low enough to require a reboot after the upgrade's complete, and just plain reboots at that, When you know that it seems an inexcusably inordinate price.
I think if you're honest, you know that you would probably be one of the first ones complaining about the "big brother" and "overreaching" Microsoft if they did something like this. I agree it would be a great idea if they reworked the Windows core so that you could install most updates without needing to reboot, but all the conspiracy nutters out there would claim that it's part of some kind of nefarious plot on Microsoft's part, because everything Microsoft does is always part of some kind of nefarious plot as far as their concerned. Someone takes a shit in a restroom a few feet further from their desk at Redmond and it's proof that they're part of some kind of secret cabal that's looking to screw over people who use Linux or some equally ridiculous explanation.
If I'm really being honest, I'd love to see Microsoft ditch the current Windows codebase and start over from scratch. Take all the decades of lessons learned and design a new OS from the ground up to take them into account. Also give them a greenfield implementation where they can do all the things they've wanted to do with Windows, but it would require a massive refactoring of code or something else that makes it untenable for one reason or another. Just like Linux, and *nix in general, is stuck in the 70s and the idea of dumb terminals, Windows (and macOS) is still stuck in the 80s and the idea that every computer being an island unto itself. Neither is true any longer, but there's not a single consumer-level OS out there that takes this into account.
I see our vatnick eel is back on the job. Working the graveyard shift over as a fourth string troll.
Anyway, gives me an excuse to clarify that I don't care if someone else designs a new OS. Apple had a great opportunity and then whiffed at the last second when they bought NeXT instead of Be Inc. BeOS at least would have brought them a little closer being that it was all-in on multi-threading when dual-core CPUs were still a hypothetical, but instead they just slapped an 80s style UI on top of a 70s era OS. Arguably it worked out for them because the iPhone and whatnot have been a license to print money for them, but it's been a huge loss for what might have been in terms of OS design. I don't care who comes up with it, Microsoft, Apple, some startup no one has heard of yet, a research project by some grad students... Even the Mach kernel idea, the cutting edge of computer science theory on the subject, was from 1985. That'll be 40 years ago next year. The newest ideas in the computer science world are decades old. Where are the Ken Thomsons or Ben Richies of the last generation? People who actually push the boundaries of what is possible and create something new? I don't give two vatnik eel sock puppet accounts who comes up with it, I just want to see someone advance the world of computing past the mid-1980s.
Like, just for one example... the whole concept of files and directories should just be done away with. Make the filesystem a giant searchable database. You shouldn't need to know, or even care, where a specific document is stored. Sort of like how you don't know where the file is physically stored on a HDD/SSD, that abstraction should carry through to the UI level. You open a word processor, it gives you a listing of documents, you pick the one you want, and away you go. At the filesystem level it might be there's a table for MS Word documents, and every Word document is an entry in that table. You wouldn't really even need specific file formats anymore, either, everything would just be plain text or some kind of binary encoding. AI could actually be somewhat useful here in processing spoken search requests like, "Open the document I was working on last Friday about rollercoasters." The system pulls up every file modified on the requested date, looks for any that have "rollercoaster" in it, and there you go. Even if you're a professional rollercoaster reviewer, and almost every document you create has the word "rollercoaster" in it, the date would probably limit the number of hits to a very manageable number. By the same token, you shouldn't really need to know or care where a file is stored. Is it local, is it on some server half-way across the world, is it at some data center the next town over? Who cares!? Setting aside the fact that probably no one besides you cares about your collection of gay midget fireman porno scripts, you can just encrypt everything to address privacy concerns. Surely someone, somewhere out there, has some ideas on how we could improve the "modern" GUI past the same concepts demoed by the people who created the GUI.
You forgot to mention, that was the first version of Windows that introduced a background defragger. Because you know how prone NTFS is to getting fragmented, and slowing the system down?
This introduction just happened to coincide with the growing popularity of SSDs. And many of those SSD users discovered that Vista was shortening the life of their drives, with its continual rewriting of sectors. So one of the many fixes made in Windows 7 was to check if the drive was an SSD, and skip running the defragger on that if so.
Meanwhile, the Linux users were laughing, because their developers knew how to create filesystems that were less prone to fragmentation.
Meanwhile, the Linux users were laughing, because their developers knew how to create filesystems that were less prone to fragmentation.
WTF are you talking about!? While I can't claim to know everything there is about the likes of ZFS and some of the others, Ext2/3/4 is every bit as prone to fragmentation as NTFS or FAT. It's just there's not really any particular need, outside of warm fuzzy feelings, to waste time defragmenting for about 99% of use cases. If you're doing video editing or something where you have large files that are being read sequentially, sure, it might be worth the effort. For your average office worker, or home user, it's a waste of time.
The only real thing Ext2/3/4 has over NTFS or FAT is that it will first look for an open space large enough to fit the entire file. failing that it will try to split it into as few chunks as possible. This actually creates additional overhead, because it's not like these searches are free. You can quibble over whether it's less overall, but honestly we're talking about segments of time below human perception so it's an academic discussion at best. If you really wanted to be serious about this sort of thing, you'd take angular velocity into account and put the most frequently accessed files on the outermost tracks of the drive, and the least frequently used ones on the inner most tracks. That, of course, would require near constant shuffling of files. which would create a lot of additional overhead, so no one is going to do it unless it's one of those rare use cases where even tiny fractions of a second are important.
You really need to get off your high horse. You're not special because you use Linux, especially people who seek out distributions that look and act as close to Windows as possible. Linux is not some magical OS where everything is sugar and rainbows all the time. It's a good server OS, a decent embedded OS, and a mediocre (at best) desktop OS. Linux was intended to be a Unix-like OS from the very beginning, and Unix, as you may or may not know, was always intended to be a server OS. Later it also became a workstation OS, but it was never really intended to be used by your average home user. You can shoehorn that sort of functionality into it, sure, but it's always going to be an ugly hack because the OS was just not designed with that sort of use in mind. It continues to be the case that desktop use is, at best, an afterthought when it comes to kernel development. If desktop functionality benefits as a result of something else they add, bonus, but it's not their primary intention. It never has been, most likely never will be. All the commercial interests that are actually doing most of the work on the Linux kernel these days are interested in it as either a cheap and customizable server OS or a cheap and customizable embedded OS. They don't give a wet raccoon's fart about Linux as a desktop OS. It was designed around the way the computing world was in the 70s and 80s, and has continued to live in that time period ever since.
Everyone goes through the honeymoon phase with Linux, I did too, then eventually most of us mature past it and realize that the OS is far from perfect. There are plenty of things it does very well, plenty of others it does extremely poorly, and plenty of ways Windows and/or macOS are so far beyond it, it's not even funny.
"If you really wanted to be serious about this sort of thing, you'd take angular velocity into account and put the most frequently accessed files on the outermost tracks of the drive, and the least frequently used ones on the inner most tracks."
If you want to get into track optimisation you put the most accessed material as close together as possible to minimise head movement between it. To minimise head movement to the less often used material you put that most used material in the mid-range tracks so you seldom have to do a seek right across the width of the disk.
One of the issues with multi-user computers is that files are very rarely loaded in single, large operations. Too often, the fact that many processes are running apparently at the same time means that loading a file, any file, is often done in chunks as other programs also make I/O requests.
As a result, it is less important to have files in contiguous disk sectors, and more important to order the disk operations from all programs to minimise head movement (back with moving head spinning disk drives).
BSD UNIX introduced what was called the 'Elevator Algorithm' which sorted the pending disk operations so the heads moved from outer to inner tracks in sequence, then back from inner to outer again. This minimised the distances between track seeks, increasing throughput and reducing wear on the drive.
Indeed, back when I first started administering UNIX systems, and when disks and disk-bus speeds were much slower than now, there was a test program that would measure the read and write speed, and the transfer speed to the system for a particular drive, and work out how to interleave the numbering of the sectors during a format operation so that by the time a sector was read, and then transferred (while the disk was still moving under the head), the next numbered sector would be under the head ready to be read or written. We are spoiled nowadays by whole track buffering within the drive electronics, DMA transfers and multiple disk command queuing within the interface hardware and drive controllers. This is even more so with solid state disks where track seek time and rotation speed are non-existent.
"Unix, as you may or may not know, was always intended to be a server OS"
Ah, but you're missing one important fact. When UNIX was designed, there was no concept of client/server computing. Computers were large, monolithic systems that if you were lucky, allowed many people to use one concurrently. And UNIX (which was run on not so large systems) was designed, almost from the outset (sometime around Edition 4 in the early '70s) to be a multi-tasking, multi-user operating system. People interacted with it much as they do with their PC's now (minus the GUI, of course, they came later) by using serial connected terminals. But the important fact is that people used UNIX systems for their daily computing needs. Mail, document preparation, code development, gaming, the whole gamut.
In the '80s, before Windows was really a thing, and well before Linux, I worked at a part of AT&T where we had Amdahl mainframes running UNIX as our main computing environment. Everybody there used UNIX, right down to the secretaries.
I was working supporting an Amdahl 5890E running UNIX, and we frequently had 200+ concurrent users on the system (although other locations used even larger Amdahl and IBM systems), doing all the things that users do now (we even had a spreadsheet called 20/20, and also had UNIX systems running Uniplex). And this even included GUI's. We had 5620 and 630 Blits allowing a number of graphical things, although I admit that much of this was diagram production with some WYSIWYG document preparation. There was a proto-IDE available apparently, but I didn't see it used.
And before that I used Edition 6 and 7 on PDP11's, shared with a community of dozens of other people at a number of educational environments, both as a student and later an administrator.
So to claim that the roots of Linux came from servers is about as wrong as you can be. UNIX always was a user-centric operating system (as many, many students in the late '70s and '80s at a lot of universities will attest), which also became as good a server OS as anything else on the market at the time.
Ah, but you're missing one important fact. When UNIX was designed, there was no concept of client/server computing.
No, that was precisely my point. *nix is stuck in the 70s design-wise with clients being dumb terminals connected to a powerful server, Windows and macOS are stuck in the 80s with each computer being a self-contained ecosystem. Instead of recognizing this simple fact, both camps just put their fingers in their ears and yell, "lalalalalalala! I can't hear you! lalalalalalalalala!" Asshole evangelists who just need to see a fucking therapist to deal with their esteem issues, have perpetuated this problem by turning it into some kind of holy war where you're either part of their blessed tribe of the chosen ones, or the unclean heathens. It's caused nothing but stagnation in the development of operating systems for at least the last 30-years.
"But, even on sites like this, which cater to more technically inclined users, rarely do you see anyone who has taken even basic steps towards debugging the issue to show that it really is the OS."
If the upgrades to the OS, made by the OS vendor who has charge Real Money for the OS (albeit via the H/W vendor) crash because, as you say, they've cut QA costs there there's no reason to look beyond the OS and its vendor.
We've seen a fairly recent example where the upgrade required a recovery partition larger than the default installation recovery partition and would either hang at about 20% complete or throw an error. And the vendor's advice to fix the "everything can be done with clicks" OS was to drop down to the command line, resize a partition. drop and recreate a partition (and hope the command-line newbie would drop the correct one) and then do some configuration based on information they'd recorded before the partition drop. W2K was a reasonably effective OS. The above is the reason a lot of us think it's all been downhill from there.
I bought a new laptop with Vista pre-installed, a Toshiba Core Duo rig. Out of the box, with no apps installed by me, just the latest MS updates, the video display showed "tearing" artifacts when I moved the mouse "too quickly" across the screen. I re-formatted the drive in anger and installed Ubuntu, and used it as that laptop's OS for years, until the hardware itself died. I was able to play the Windows version of "Battefield: 1942" on it via WINE.
Legend has it that NT 3.5 was probably the least buggy version of Windows ever because they were going to release it alongside the 486, but then Intel had to delay release by like 6-months leaving the Windows developers nothing to do but sit around refining the code.
I actually liked Windows 3.1, except for the GPF errors and the like that would come out of nowhere because one app would clobber the memory space of another. If they had just made a more consumer oriented version of NT 3.5, that would have been awesome, but of course they didn't want to break that DOS compatibility and most PCs at the time didn't really have the resources to be able to run copies of DOS in a VM.
PCs of the time were just fine running multiple copies of DOS in a VM, they were just using OS/2 to do it instead of NT 3.51, and needing less memory. It's possible Desqview and Desqview/X could also have been options, but I've never used those.
OS/2's VM capability isn't perfect, but is pretty decent and reasonably fast. NT's DOS capability isn't quite as horrific as you might recall, but is quite considerably slower (twice as slow or more than a DOS VM under OS/2 in some cases, on the same machine).
Whilst installing various OS/2 versions and NT 3.51 on the same 486 PC I had the chance to compare them. They're early to mid nineties OS, so really nothing comes out smelling of roses. OS/2 has various neat features that work to different degrees - interface, compatibility, scripting,multimedia. NT had solid win16 support, good networking, OpenGL that largely just worked (and 3D accelerated in some cases which is more than OS/2 ever managed) - DOS was slow though, good luck if you didn't know the installation requirements back to front as it'll happily fail with no apparent errors, and a very basic interface.
NT 3.51 looked OK on the surface and mostly worked - it was a lot better than Windows 3.11... But as someone who wrote image processing code that maxed out the CPU, memory and hard drives it quickly became clear that it was half-baked.
+ A lot of the advertised features (POSIX subsystem - haha) were missing all together or simply half implemented.
+ The virtual memory system & *file* caching were plain wrong - poor bad design choices meant that NT would cache sequentially read (and written) files in RAM, and *page out* your application's working-set to disk in order to accommodate all those never to be used again bits of file. Essentially NT would put your files in memory and your application on disk. Dumb as rocks.
+ Perhaps Microsoft did foresee the problem with file caching - because they provided some methods to flag opened files as being sequentially accessed, however those calls did nothing in practice, NT still paged out your application to disk so it could keep files (that were much bigger than the available RAM) in memory... Sigh...
By contrast Linux of the same era was a *lot* quicker for that application simply because it made more sensible choices with respect to balancing the demands of the block cache vs the application's working set. I was able to take the highly tuned "NT specific" code - naively convert it to POSIX and run the application much *faster* under Linux on the exact same machine. Furthermore the performance did not degrade when the image file sizes exceeded the available memory (unlike NT where the machine basically thrashed the drives to bits until folks got bored of waiting for it and pulled the plug).
Until you mentioned how "good" Linux was at the time. The 1.0 release? Really. Now 2.0 which became usable just around when NT 4 was shipped was a maybe. If you could find a full set of stable device drivers. Not an easy task. And unless you were a VAR selling a complete package a total non starter for end user software.
NT 3.51 was 1995 to 1997. When 4.0 came out. NT 3.51 was rock solid. NT 4.0 was back to BSOD land. The Win32 subsystem on NT 4.0 was so different in behavior from Win 95 that you had to develop on Win 95 boxes because code that worked on NT 4 would fail on Win 95. A tradition that MS maintains to this day across their "compatible" OS platforms.
Now the funny thing was I was also working on high end image processing / video processing software at the time (VFX packages for Hollywood movies type stuff). From scratch. On the Mac because Wintel just did not have the horsepower. And I ran into none of the type of problems you seemed to have run into. And we maxed out all parts of the hardware. Our biggest problem was making sure all the IPUs in the CPU were full every clock cycle and the cache miss hierarchy was properly optimized. Regs/L1/L2/storage etc. We were at the clock-cycle counting level of optimization. In asm. It made a big difference when you were compositing graphic ops on huge images real-time.
Sounds like you were porting a Unix box package to NT (SGI?) and did not have enough target platform experience to know that most MS API's sorta worked, kinda, but you had to QA all API's to see which documented features worked and which ones did not. Even basic stuff had serious crash bugs. Like even tooltips. The MS POSIX stuff was a sloppy port from another project so of course was about as reliable as a undergrads end of term project. Welcome to the wonderful world of Redmond.
If you want to see the NT3.5.1 / NT 4.0 source its all been out in the wild since 2003. The full build minus most of the network stack. The kernel code (a.k.a the DEC code) is really nice.The rest is a total dogs dinner The Win2K/XP source code is also out in the wild. Pretty much the same as the NT4 code bases. But with more crap on top. If you want a real good laugh look at the WinExplorer source code. Whoa. You will never see code that terrible. Anywhere. Ever.
As I said, welcome to Redmond. Or in the above case. Just across the border in Bellevue.
"POSIX subsystem - haha"
Not that ha-ha ... it was just enough to allow Government contracts requiring compliance with FIPS 151-2. Made a lot of people quite a bit of money.
More important to me, it also allowed Cygwin to exist to the extent that it did, which made the NT command-line quite a bit more useful than it would have otherwise been .... almost to the point of Win2k being a decent platform. Almost.
This post has been deleted by its author
Wiin2K sever is still on a factory machine here. I do like the interface better than Win10, or any of the BSD / Linux machines we have, but it's also had the advantage that it's really easy to port to new hardware.
Win2K only has enough resources to add maybe one new set of unused motherboard drivers (WinXP is better), but swapping the disk drivers and kernels is easier, and when the USB ports fail (a typical hardware fault), 45 minutes to add the new drivers and pop the HD into the new machine. Boot, clone the HD to a new HD, and ready to resume production.
Win2k was clean and didn't yet have the XP or Vista drapery. There was less focus on bells and whistles and more on functionality, and; it was the original AD driven environment.
Win2k was the best (from M$) period done. :)
Now at the same time I was also completely enamored with slackware and it's simplicity
Mr Speed, I wholeheartedly agree. NT 3.51 is my favorite Microsoft OS.
Win7 was arguably the best MS OS ever, and the Hotpatch feature available in Windows Server 22/25 elevates it, bringing a long, long needed improvement to managing Windows servers.
The massive upgrade NT 3.51 was from everything before it (NT 3.5 especially inclusive), the refinement of the interface, everything about it was just beautiful.
There's a methodology for designing, building and bringing up complex software systems and it doesn't involve heaping everything together, hitting the 'go' button and then trying to make sense of reams of debug messages, dumps and what-have-you when the software eventually falls over (followed by piecemeal fixing code in the order of problem obviousness). Also, just because the code didn't crash, or "ran overnight" or "ran over the weekend" doesn't mean its working.
However, this kind of "hose the code at the wall and see what sticks" software debugging is also prevalent -- I've talked to innumerable programmers over the years about this and the general impression is that talking to a brick wall might be more productive. This seem partly because, like everything else computer related, the boundary between 'obvious' and 'incomprehensible' is vague and partly because management wants results, if not this morning definitely by the end of the day.
>> The two sat before the CRT attached to Plummer's 486DX2-50 development machine and fired up the code. It took a while – log messages scrolled past on a VT220 terminal connected to the serial port – but the debug build did its thing.
I was more amazed that all their Dev team had were relatively humble 486s after 1995. What did they compile code on: Pentium 60MHz with the FDIV bug in?
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
