I thought Netware was pretty good for what it was... not nearly as powerful as a UNIX, but that's not what Netware shops usually needed. I administered it next to UNIXes (and some NT sites, ugh) in my youth when I jobbed as a system administrator before studying, and have mostly good memories of it. It didn't cause troubles and served a small to mid-size company well. Fond memories...
And be born more than 25 years ago and had access to these computers.
Even though I'm old enough to have used DOS as a kid, my grade school didn't have any computers for student use. I jumped immediately to Windows 95 from DOS, but never used it for more than playing games and Microsoft Office.
I encountered macOS 8 in High School but only used them for Logo (turtles!).
What’s most amazing to me (as somebody who’s tried more beer styles than operating systems) is that the comparison doesn’t really make use of actual beer styles. There’s just the suggestion that Amiga is now an “import” beer.
It’s a good reminder that >25 years ago most bars and grocery stores has just a few different brands of larger, and maybe a light option or two.
In the mid-90's Heineken and Becks were considered "fancy" beers. By the late 90's , microbreweries were starting to appear (at least in the Chicago-area where I was living).
The thing that preserves it is that the beer style that tastes like licking a Christmas tree but has a high ABV won out. People still use Windows for the same reason.
I guess this is a good place to ask. Back in the 90s there was a humor mailing list that was like, "Top Five such-and-such" and it was operated by someone named Chris. Don't remember last name. Does anyone remember this?
Linux beer: Displaced the various flavors of UNIX beer by being free (as in beer) and is extremely popular where that's important, i.e. everywhere. Tasted a bit manky in its earliest brewings but eventually got tasty enough to displace all the other UNIX beers, although Solaris beer brewers fiercely insist to this day that it's not nearly as good. Currently the most widely drunk beer in the world, although its detractors point out that for most of that it's only an ingredient and not being drank straight up. Its drinkers have a one-sided rivalry with Windows beer drinkers for some reason even though the latter has long since ceased to care.
Nowadays, the pop-tops don't break off as often - though when they do, you still either need a complete set of instructions, or a friend who has been drinking Linux Beer for several years. Also, available in cans of every imaginable size in every imaginable bar.
AmigaDOS (and Kickstart & Workbench) have always been 32 oz variety. Even Motorola 68000 got 32-bit address space and 32-bit register width. Same binary code runs fine on MC68020 and beyond.
Internally MC68k was 16-bit, but from a programmer's point of view you got 8x 32-bit data registers, 8x 32-bit address registers (minus stack pointer) and a flat, 32-bit memory space.
No 32-bit x 32-bit = 64-bit multiply though, only 16b x 16b = 32b. Same with divide.
Yeah, that upper 8-bit address tagging thing was a "clever"... waste of cycles.
Just rotating/shifting 32-bit registers by 8 places costs 24 cycles (8 +
2 * n) on 68000, like if the value was passed in a register. Shifting the value by 24 places is even worse, massive 56 cycles...
You could also do "move.b (a0), d0" directly from the pointer to pointer address. That'd cost just 8 cycles.
Self-modifying code wasn't all that beneficial nor popular on 68k because of ample number of usable registers, 16, including a7, that is used as a stack pointer, but nothing prevented you from using it for something else for a while. I think copy protections and such used weird tricks like SMC.
Machine code generators (kind of limited JITs) were useful, but those usually generated large enough blocks of code to flush a small cache by default.
There were also some race conditions with DMA (like blitter and perhaps with disk), when DMA completion wasn't properly checked.
But 99%+ of code didn't use those tricks and was perfectly compatible.
I got a good chuckle out of an old "which os are you" quiz (html site, pre social media quizzes) where I ended up Gnu/Hurd because I forgot to answer one question :D
iOS beer: Everyone drinks this! Your grandma, kids at school, the guy driving you to the airport. Most people don't even know they're drinking beer until they're too impaired to perform basic tasks. The makers of iOS beer believe the solution is simple: drink more iOS beer!
Depends on what you mean by "good OS". Architecturally, it was a descendant of VMS and its kernel was considered well engineered for its time. It had a ton of bugs and security issues but so did all the pre-2.0 Linux kernels/distributions that were its contemporary.
From the userland and UI perspective, well, it's just Windows, both in the positive and negative sense; NT is the granddady of Windows XP and Windows 7 both of which were well regarded but NT also suffered from all the warts and ugliness caused by being converted into a consumer OS.
NT5/Windows 2000 was a huge leap forward for Microsoft in terms of design model and security. NT5.1/Windows XP had some iterative improvements and some things going backwards.
NT4 was better than 95 in every way, but was merely mediocre. Microsoft's suite of workplace tools that ran on NT4 were janky and broken, requiring their own college degree to set up and administer, but they also represented a real productivity boost for many teams.
I liked windows NT so much I traveled an hour on public transit to stand in a two hour line to get a beta copy of Windows NT 4, and ran it as my daily driver (upgrading to non-beta) until I just switched to Linux (and then the first version of MacOS X).
This was after I tried to take the NT 3.51 kernel and marry it with the Win95 UI (that didn't work).
> This was after I tried to take the NT 3.51 kernel and marry it with the Win95 UI (that didn't work).
How in C:\>'s name would that have been possible? Windows 95 was an unholy mixture of 16bit code running in vm86 mode interleaved with 32bit code, and nearly full Windows 3.1 compatibility even in lower level aspects. NT was a relatively clean full 32bit reimplementation.
Was the UI part so cleanly separated to have a chance at that? (Not what my impression was, but I really never dived that deep into Windows.)
> I was dumb and naive and just tried to replace the kernel on win95 with the WinNT kernel.
Hahah, I'd be lying if I said that I didn't do equally naive stuff back in the day, I know exactly what your kind of thinking was.
It's almost sad, because understanding takes the magic away. I sort of miss looking at computers with the eyes of a child and wondering what lies behind the abstractions...
I have an even better dumb story. Right after high school I visited my friend at Berkeley, and while there I wrote down his IP address.
When I got home I changed my IP to be a different one on the same subnet and matched the gateway address. I knew what a subnet was, I knew what a router was, but I didn’t realize you can’t just reach any router you want.
I figured if I did that I could get all the stuff that was restricted to the Berkeley network.
I got a fancy HPUX refrigerator-sized server to replace a mainframe in the early 90's. Was on Netware 3.11 at the time and did not understand the concept of IP, so just made up a bunch of random IP addresses. The confusing thing was telnet did not work at all, but ftp kinda/sorta did. That was an embarrassing call with HP support, but the guy I talked to kindly gave me a 30-minute tutorial and usenet filled in the rest later (and crynwr packet drivers).
One time, I really wanted to get an EXE of my QuickBasic programs. I just renamed the file to be myprog.exe. Needless to say, that didn't work. But little me learned that day, so I'm not gonna be embarrassed by it.
It was light and day. It is what it is because of the man in charge. Billy G knew that even he didn't have the skills to manage the development of NT. Therefore he got a guy called Dave Cutler to head it up.
Dave was the man because he used to work for DEC on a project called Onyx and prism (A hardware/software combo to do essentially what NT does) was cancelled by DEC at the last minute.
MS wanted him that much that they brought his entire Development team to MS (not one person was left behind, at his demand. They where quite ...tribal...)
This is also part of the reason NT originally worked on DEC systems as well.
It's all in a rather fascinating book called Showstopper: The Breakneck Race to Create Windows NT and the Next Generation at Microsoft
For a few bucks it really is a most excellent read that goes into some techy and really fascinating issues they encountered. One of the biggest was Memory was so little back then.
I quite like the NT kernel from a design standpoint. Sadly some of the cooler features were removed over time, like offering Win32, Posix and OS/2 APIs as equal subsystems for interacting with the kernel (though the capability became useful again with WSL1). The entire user management and rights management story is very well thought out; somewhat complicated but incredibly powerful. And while it's not the fastest kernel out there, it really excels at keeping the system interactive and useful under CPU and memory pressure.
If only the userspace was as good as the kernel. A lot of good things were (and still are) obscure because they aren't properly surfaced in the Windows UI.
There is a certain elegance in the OSs of the 1990-2000. NT was a bit slow but predictible. It got slower as you loaded it. Now, with preemptive multitasking some things are very fast and some are very slow. Why do i have to wait seconds for the cursor to blink ? Why it doesn't blink at a constant rate?
I used it to run Pro/Engineer (now called Creo) in the mid-90s. Prior to that, it only ran on Unix workstations from Sun, SGI, etc.
I also had Macs for personal stuff at the time. I remember NT being remarkably stable in comparison to regular Windows, and MacOS of the time. It would run for weeks and weeks without crashing or rebooting. It wasn't sexy, but it worked.
An innovative new taste that reminds some of iOS beer when mixed with Windows XP beer. This deafening non-distinctiveness will earn it a place in history as an exemplifier of late 2010s style. It comes in a 64 oz can which, following earlier experiments with square containers, is round again, in fact all corners have been completely removed with only one edge remaining. This 64oz can can contain unlimited numbers of other 8, 16, 32 and 64 oz cans of other beers, and the label comes in all languages except for some markets where it doesn't. In the interests of your safe drinking and continued loyalty to the beer, the can will stop you drinking from it just when you really want to while refilling itself with what you hope is not another flavour.
You really wonder why they bothered calling this an "all new beer" when it's really essentially the same beer as their last one, just with slightly better looking label and can. It still has those weird aftertastes and smells in it that you swear were from the Windows95Beer and DOSBeer.
Kind of depressing to contrast that with today where there are just three deeply entrenched operating systems (phone and embedded OSes don't count since the end-user has no meaningful access to them). We, or at least I, naively thought we'd see continued rapid evolution of a plethora of OSes to accompany the rapid evolution of computer hardware.
We were already seeing that happen back in 95. From that list half the operating systems were either deprecated or niche. We still have a plethora of niche platforms now from the various BSDs, Open Solaris derivatives, Hakiu, ReactOS, and all the experimental ones too.
The hardware duopoly was already in full force by 1995. Sure Amiga was listed in that article but by 1995 IBM-compatible PCs had won and Amiga, Atari, Acorn etc were basically just used by a small few enthusiasts or specialist domains.
Now that you mention it, indeed a pretty big omission! (For those who never dealt with it, it was a full blown operating system on the server side, with own multitasking kernel that had its own proprietary filesystem and everything.)
I can enrich the VMS beer:
Despite having Open in the name, the newer cans are tightly sealed and come in 32 or 64 cans. However these cans fit only very specific and expensive 32 or 64 oz can holders.
Some say the first sip tastes a lot like the well known DOS flavor but it rapidly changes to its unique profile.
The discussions here about Windows NT, Netware, and the various UNIXes transports me back to my youth when I worked as a freelance system/network admin for various small to mid-size companies before going to university. What a fun, diverse job that was. Very different from what I imagine DevOps to be nowadays...
It's kind of weird that Linux in a way became what people were afraid Windows would - Linux has absolutely crushed and destroyed all other OSes so that all that remains on server is basically Linux and a small amount of Windows in a bad POSIX mustache with a sign around its neck that says "LINUX".
And worldwide we're basically down to three or five OSes: Windows, macOS, Linux, Android, iOS, depending on how you count them. The various "flavors" of Linux are way more similar than even old versions of Windows were.
> It's kind of weird that Linux in a way became what people were afraid Windows would
The fear surrounding Windows was that a company would own computing even more than IBM owned computing in the 1950s: Microsoft would be it, every other company would exist at Microsoft's sufferance and effectively pay a tax in order to be allowed to keep existing, and anything which would threaten Microsoft's current business plans would be destroyed through a mix of API changes, revocation of developer licenses, AARD code-like subterfuge, and, in extremis, actual lawsuits.
Wow, did a double take seeing that. I took all of the courses he taught at Caltech around 2011-12. Really enjoyed him. Didn't expect to see that link on HN.
People did stick to 3.1, at least on 386 and 486 machines. I knew nobody who ran Windows 95 on anything less than a Pentium, even though it was compatible down to the 386 in most cases.
It's just that so many people bought their first computer around that time, that it really was able to grow quickly.
This person wasn't a Mac user; that entry is both inaccurate and uninteresting. If someone with more patience than me would like to rewrite it: From the programmer's perspective, the Mac was always essentially a 32-bit machine (you could make an argument for the early Mac being limited to 24 bits), and the "ingredients" were actually very well documented in the early Inside Mac volumes; lockdown came later as Apple tried to move the application base to a succession of "modern" OSes.
71 comments
[ 3.3 ms ] story [ 151 ms ] threadFeels like something that someone pulled from Usenet and turned into a web page, back when the internet was still interesting.
Even though I'm old enough to have used DOS as a kid, my grade school didn't have any computers for student use. I jumped immediately to Windows 95 from DOS, but never used it for more than playing games and Microsoft Office.
I encountered macOS 8 in High School but only used them for Logo (turtles!).
So I really don't get the article.
It’s a good reminder that >25 years ago most bars and grocery stores has just a few different brands of larger, and maybe a light option or two.
Something about brevity being the soul of wit.
OpenBSD would be just can, by default. Beer is your prerogative, and available in the ports tree.
Internally MC68k was 16-bit, but from a programmer's point of view you got 8x 32-bit data registers, 8x 32-bit address registers (minus stack pointer) and a flat, 32-bit memory space.
No 32-bit x 32-bit = 64-bit multiply though, only 16b x 16b = 32b. Same with divide.
One "clever" trick was to store data in the highest byte of 32-bit pointers — which had worked on the 68000 because it had only a 24-bit address bus.
Another thing that didn't work was self-modifying code. The 68020 introduced an instruction cache, which would have had to be cleared first.
Just rotating/shifting 32-bit registers by 8 places costs 24 cycles (8 + 2 * n) on 68000, like if the value was passed in a register. Shifting the value by 24 places is even worse, massive 56 cycles...
You could also do "move.b (a0), d0" directly from the pointer to pointer address. That'd cost just 8 cycles.
Self-modifying code wasn't all that beneficial nor popular on 68k because of ample number of usable registers, 16, including a7, that is used as a stack pointer, but nothing prevented you from using it for something else for a while. I think copy protections and such used weird tricks like SMC.
Machine code generators (kind of limited JITs) were useful, but those usually generated large enough blocks of code to flush a small cache by default.
There were also some race conditions with DMA (like blitter and perhaps with disk), when DMA completion wasn't properly checked.
But 99%+ of code didn't use those tricks and was perfectly compatible.
https://web.archive.org/web/0/http://www.somethingofthatilk....
Another site bites the dust. A gallery of some of the comics, if archive.org is slow: https://imgur.com/gallery/mukhL
Everyone drinks it. It's non-offensive and ubiquitous, but doesn't offer very much flavour.
(Heavy iOS user here)
Was Windows NT a good OS? My professor in University was a hardcore Linux fan, but he used to praise NT saying it was the best thing Microsoft did.
From the userland and UI perspective, well, it's just Windows, both in the positive and negative sense; NT is the granddady of Windows XP and Windows 7 both of which were well regarded but NT also suffered from all the warts and ugliness caused by being converted into a consumer OS.
NT4 was better than 95 in every way, but was merely mediocre. Microsoft's suite of workplace tools that ran on NT4 were janky and broken, requiring their own college degree to set up and administer, but they also represented a real productivity boost for many teams.
Microsoft has always had a shit ton of extremely smart people, mired in two shit tons of middling execs.
Windows NT was what happened when the product / political side of the company ignored the technologists for a bit.
This was after I tried to take the NT 3.51 kernel and marry it with the Win95 UI (that didn't work).
How in C:\>'s name would that have been possible? Windows 95 was an unholy mixture of 16bit code running in vm86 mode interleaved with 32bit code, and nearly full Windows 3.1 compatibility even in lower level aspects. NT was a relatively clean full 32bit reimplementation.
Was the UI part so cleanly separated to have a chance at that? (Not what my impression was, but I really never dived that deep into Windows.)
After booting it just sat there like “what in C:\>’s name have you done to me?!”
Hahah, I'd be lying if I said that I didn't do equally naive stuff back in the day, I know exactly what your kind of thinking was.
It's almost sad, because understanding takes the magic away. I sort of miss looking at computers with the eyes of a child and wondering what lies behind the abstractions...
When I got home I changed my IP to be a different one on the same subnet and matched the gateway address. I knew what a subnet was, I knew what a router was, but I didn’t realize you can’t just reach any router you want.
I figured if I did that I could get all the stuff that was restricted to the Berkeley network.
Needless to say it didn’t work.
Dave was the man because he used to work for DEC on a project called Onyx and prism (A hardware/software combo to do essentially what NT does) was cancelled by DEC at the last minute.
MS wanted him that much that they brought his entire Development team to MS (not one person was left behind, at his demand. They where quite ...tribal...)
This is also part of the reason NT originally worked on DEC systems as well.
It's all in a rather fascinating book called Showstopper: The Breakneck Race to Create Windows NT and the Next Generation at Microsoft
For a few bucks it really is a most excellent read that goes into some techy and really fascinating issues they encountered. One of the biggest was Memory was so little back then.
If only the userspace was as good as the kernel. A lot of good things were (and still are) obscure because they aren't properly surfaced in the Windows UI.
• It is old enough where it has to explain email to the reader, and explain why Gates would be obsessively looking at it throughout the day.
• The project got slowed down by adopting a cool but unproven new language called C++. The 'rewrite it in Rust' of its time.
I also had Macs for personal stuff at the time. I remember NT being remarkably stable in comparison to regular Windows, and MacOS of the time. It would run for weeks and weeks without crashing or rebooting. It wasn't sexy, but it worked.
An innovative new taste that reminds some of iOS beer when mixed with Windows XP beer. This deafening non-distinctiveness will earn it a place in history as an exemplifier of late 2010s style. It comes in a 64 oz can which, following earlier experiments with square containers, is round again, in fact all corners have been completely removed with only one edge remaining. This 64oz can can contain unlimited numbers of other 8, 16, 32 and 64 oz cans of other beers, and the label comes in all languages except for some markets where it doesn't. In the interests of your safe drinking and continued loyalty to the beer, the can will stop you drinking from it just when you really want to while refilling itself with what you hope is not another flavour.
The hardware duopoly was already in full force by 1995. Sure Amiga was listed in that article but by 1995 IBM-compatible PCs had won and Amiga, Atari, Acorn etc were basically just used by a small few enthusiasts or specialist domains.
And worldwide we're basically down to three or five OSes: Windows, macOS, Linux, Android, iOS, depending on how you count them. The various "flavors" of Linux are way more similar than even old versions of Windows were.
The fear surrounding Windows was that a company would own computing even more than IBM owned computing in the 1950s: Microsoft would be it, every other company would exist at Microsoft's sufferance and effectively pay a tax in order to be allowed to keep existing, and anything which would threaten Microsoft's current business plans would be destroyed through a mix of API changes, revocation of developer licenses, AARD code-like subterfuge, and, in extremis, actual lawsuits.
https://en.wikipedia.org/wiki/AARD_code
Add Palladium on top of that, and PCs are better locked-down than cell phones turned out to be:
https://en.wikipedia.org/wiki/Next-Generation_Secure_Computi...
and the hilarious "if programming languages were cars" list: http://users.cms.caltech.edu/~mvanier/hacking/rants/cars.htm...
Wow, did a double take seeing that. I took all of the courses he taught at Caltech around 2011-12. Really enjoyed him. Didn't expect to see that link on HN.
It's just that so many people bought their first computer around that time, that it really was able to grow quickly.