Most obscure/upvoted answer will get the answer vote!

Most Unix and Unix-like systems have a "traditional history" that is passed down with each generation, so that newcomers can see and understand some of the more esoteric portions of the system. What can the long-time Windows admins out there contribute as to the "Legendry" of Windows, as far as esoteric knowledge that new admins should be exposed to?


There is a widespread urban legend in the Windows community that Dave Cutler chose the name "Windows NT" (WNT) because if you subtract one from each letter you get VMS, which is the operating system that Dave Cutler worked on before he came to Microsoft.

However, that is not true. Neither did Cutler choose the name (it more or less emerged by accident) nor was the name WNT (it was, in fact, "NT OS/2", because at that time, the idea was still to build the next generation of OS/2 – Windows wasn't even in the picture). NT also doesn't stand for "New Technology".

Dave Cutler was hired by Microsoft to build a next-generation operating system for the PC. At that time, Microsoft's most modern OS was OS/2, built jointly by IBM and MS. And so, to MS, "next-generation" naturally meant a new version of OS/2. (Windows was just seen as a GUI for DOS, that would soon be replaced by a real OS.)

At DEC, Dave Cutler had worked on VMS. One of the problems with VMS was that a lot of it was written in assembly, and that it was fairly monolithic. Both meant that it was somewhat hard to port to other platforms. When he was tasked to write a successor to VMS, he changed all of that; while he liked the abstractions and style of VMS, he didn't like the implementation: he wanted to write it in a highly modular style, and he wanted to do it in C – basically a cleaned up and modernized re-implmentation of VMS. DEC soon lost interest in his project, and effectively mobbed him out of the company, which is how he ended up at Microsoft, where he was able to implement his ideas. (Which is why NT looks so similar to VMS: it's essentially the version of VMS Cutler always wanted to write but DEC didn't allow him to.)

So, Cutler was writing a new version of OS/2. At that time, it was not at all clear, what exactly a "PC" was and that the i386 would win out. Intel was experimenting with RISC processors, Microsoft was experimenting with MIPS processors, IBM was doing the PS/2 and later also the PowerPC, DEC's Alpha was around the corner. Building an operating system for only one of those platforms would have been a huge and dangerous gamble. So, the new OS/2 would have to be highly portable. In order to ensure maximum portability, Cutler tried to find the strangest, most PC-unlike architecture possible. And he found the Intel i860. The i860 was 64 Bit, the i386 32 Bit. The i860 had plenty of registers (32), the i386 only 8. The i860 is RISC, the i386 CISC. The i860 has a VLIW architecture. It had a SIMD unit on the chip (that would later become the inspiration for the MMX SIMD instructions on the Pentium MMX). It exposed its pipeline to the software, so that all the scheduling decisions that the i386 makes on the chip, would have to be made by the compiler (this was the inspiration for the EPIC architecture of the Itanium). And it had one of the strangest memory models ever.

In short, it was the perfect target for a highly portable OS: if you developed and tested on the i860, you could port it almost anywhere. So, that's what Cutler did.

There were two versions of the i860: a cheaper, simpler one called the XR and a faster one for SMP machines called the XP. Except MS was working with early prototypes and the marketing names hadn't been chosen yet, so they were simply known by their Intel codenames: N10 and N11.

The operating system that Cutler was working on, was codenamed Portasys, because of its portability. However, internally they didn't call it that: they were building it for the N10, and they were building a version of OS/2, so in conversations they simply called it "OS/2 for N-Ten" or "N-Ten OS/2", and because programmers are lazy, that quickly got shortened to "NT OS/2".

Then, two things happened: MS hadn't actually bothered to tell IBM that they were building their own version of OS/2. When IBM became aware of the fact that the new OS/2 was actually quite different from the old one, IBM and MS had a falling out and the joint development of OS/2 ended.

The second thing that happened was that Windows 3.0 took off beyond everybody's wildest dreams.

That's when MS changed directions 180°. And now the modular design of NT OS/2 came in quite handy.

In NT, there is a clear distinction between the kernel and what is called a "personality". Usually, the kernel's job is to abstract the hardware and present an abstraction to userland. In NT, these responsibilities are split: the kernel abstracts, and the personality presents the abstraction to the userland. Well, actually, the personality does not present the abstraction, it presents an abstraction; there can be multiple personalities.

At this point in our story, NT looked like this: there was the NT kernel, on top of the NT kernel was the OS/2 personality, and on top of that was the OS/2 Presentation Manager (GUI). The filesystem was HPFS. Because of the personality abstraction and the ability to run multiple personalities at the same time, it was now very easy to keep the system running while at the same time adding a Windows personality. Once the Windows personality was complete, Presentation Manager running on top of OS/2 was replaced with the Windows UI running on top of Windows. HPFS was renamed to NTFS and some tweaks were made.

And then some marketing guy somewhere decided that that "NT" moniker was actually kinda cool, but the "OS/2" thing had to be replaced with "Windows" and it sounds better the other way around ("Windows NT" instead of "NT Windows"). And when he was asking what NT stood for, they told him that it didn't actually make any sense because they weren't doing an i860 version any more, and so he retroactively created the backronym "New Technology".

That's how it became WNT. And, as you can see the letters "N" and "T" were chosen by chosen by Intel, the letter "W" was chosen by fate and the ordering of the three letters was chosen by some marketing guy, so there is simply no way that anyone could have consciously chosen them in relation to VMS.

And, by the way: Arthur C. Clarke has always said that he never thought about "IBM" either, when he came up with the name "HAL". In fact, he said that if he had recognized that relationship he would have changed HAL's name out of respect to IBM with which he had very good working relationship.

  • 1
    I have issue with the "HPFS was renamed to NTFS" statement. AFAIK, NTFS was a "from the ground up", though some inspiration (not code) was taken from HPFS. Sounds like you've read "Show Stopper! The Breakneck Race to Create Windows NT and the Next Generation at Microsoft", by G. Pascal Zachary (amazon.com/Show-Stopper-Breakneck-Generation-Microsoft/dp/…). Jeff Atwood recommended it on his blog at one point: codinghorror.com/blog/archives/000060.html I heartily agree-- it's a good book. Dave Cutler sounds like a great guy, and a real badass... >smile< – Evan Anderson Jul 8 '09 at 2:59
  • 1
    So Windows NT literally has multiple personalities...that explains so much. Excellent writeup. – Matt Simmons Aug 13 '09 at 20:43
  • lol BTW: the term "personality" for that kind of thing is standard. It is also used in Linux for example, which in the '90s added personalities for all sorts of proprietary Unices (e.g. SCO) so that you could run your existing apps without recompiling. Now that it is obvious that Linux has "won", the proprietary Unices do the same (AIX 5L, where the "L" stands for "Linux", HP-UX 11i, where the "i" stands for "integration with Linux"). Also, most BSDs have a Linux personality and in Dragonfly BSD even the Dragonfly part is a personality (similar to NT). – Jörg W Mittag Aug 15 '09 at 8:55

I don't have any stories personally, but I can point to some:

  • 4
    And Larry Osterman's blog @ blogs.msdn.com/LarryOsterman, too, but apparently my account here is too new to post more than one link... – BryCoBat Jul 7 '09 at 20:36
  • Give it a few more points and you'll post. – Avery Payne Jul 7 '09 at 20:39

I always thought the fact that you could change NT Workstation 3.51 to NT Server with a registry change was pretty cool. And says everything about Microsoft's market segementation strategies.

  • +1 for some lore long forgotten. I can personally verify that this existing - I remember reading about it. – Avery Payne Jul 11 '09 at 20:37

How about Osterman's explanation of ctrl-alt-delete in his post Why is Control-Alt-Delete the secure attention sequence (SAS)?

  • +1 excellent example! – Avery Payne Jul 11 '09 at 20:37

Windows has a legacy that dates back to before the first versions of MS-DOS to CP/M, including:

  • 8.3 filenames
  • Using letters to identify drives
  • Programs used to be COMmand files, not EXEcutables (*)

(*) Trivia: DOS COM files get loaded into memory at 0x0100:0000 (segment/offset); CP/M COM files got loaded into memory at 0x0100

Correction 9 July: CP/M COM files got loaded into memory at the absolute address 0x0100. DOS COM files get loaded at offset 0x0100 in the next available segment. (Thanks to Euro Micelli).

  • 1
    Not quite. COM apps in DOS (and in Windows!) get the bits loaded at CS:0100, where CS (and DS, SS and ES) was set to the next available 16 byte boundary available in RAM. COM programs are presumed to not know anything about segments and can operate within their private 64Kb window. The first 0x100 bytes of the segment are a structure called "PCB" (I think?) used to communicate command line parameters and other info to the program, and is based on (and backward-compatible with) a similar structure used by CP/M. – Euro Micelli Jul 7 '09 at 22:03
  • 1
    You're talking about something that I did entirely too much of... >smile< x86 assembler under DOS! The first 0x100 bytes are the PSP-- program segment prefix. You're correct in that the command line parameters went in here (hence the limit of 127 characters), as well as pointers to the environment and other structures. COM programs could use as much of memory as they wanted, but the binary could only be 64KB. EXE files, with their relocation table and the OS loader, allowed for executables larger than 64KB. – Evan Anderson Jul 8 '09 at 2:57
  • I stand (err, sit) corrected. I'll update the answer for those who skip reading comments. – Bevan Jul 8 '09 at 21:28

The (In?)famous 49.7 day stability bug in Windows 95. Under certain circumstances, you'd want to reboot your machine every 49 days to avoid the bug from tripping up parts of your system. No, it didn't force a shutdown or anything like that, it just produced strange results...

Some quick links found...

CNET: Windows may crash after 49.7 days

MSDN: System.Timers.Timer and System.Threading.Timer bug after 49.7 days!

The Joel on Software Discussion Group: The Windows 49.7 days shutdown bug


A couple bits of trivia:

  • Dave Cutler was one of the designers of DEC's VMS operating system, and then the chief architect of Win NT at Microsoft. Just as HAL (from 2001: A Space Odyssey) was said to be one letter before IBM, WNT (Windows New Technology) is one letter after VMS.

  • We all know what NT stands for, and there was that annoying line on the startup screen of Windows 2000 that said "Built with NT Technology," i.e. with New Technology Technology.

  • This isn't so much Windows trivia, but it's also DEC-related. DLT and LTO tape formats are descended from Digital's infamous TK-50 tape cartridges. Infamous, because TK was widely said to stand for "Time Killer." The later TK-70s were a bit better, but not much.

  • The WNT >> VMS story is apparently an urban legend. See Jörg W Mittag's story above. – Abel Nov 5 '09 at 19:41

I remember a fight with Digital Equipment way back in the day trying to get the Alpha NT linker to work to link Pro/Engineer on NT 3.5. This particular release of Pro/E broke the 32MB executable limit (wow, remember when that was a big deal)? Needless to say, I tell myself today that DEC's refusal to fix that problem for three months contributed to the ultimate demise of AlphaNT, since we'd already shipped Intel and MIPS versions, and Pro/E was THE NT app of the day.

Yes, I'm delusional, I'll admit it, but those were good times. 32MB of RAM and 1GB hard drives.

  • I own a DEC Multia, so I hear you. ;) +1 for remembering the "other flavors of Windows NT". – Avery Payne Jul 7 '09 at 20:41
  • 1
    I bought a DEC AlphaStation 2nd hand that had a release candidate build of Windows 2000 for Alpha installed when I acquired it. I seem to remember it being RC2 or RC3. The final build was never released, of course. – Clint Miller Jul 7 '09 at 21:39
  • 1
    I remember throwing away about a dozen Avantis when I moved about 5 years ago. I shed a tear at the lost potential of the Alpha line. Ironic, isn't it, HP combined with Intel to build Itanium to compete against the Alpha, and 10 years later end up owning the Alpha. I remember beta testing PPC NT 4.0 on some IBM rs6000s. That was a platform I was sad to see never materialize. – Chris K Jul 8 '09 at 15:38
  • 1
    Fortunately, the Alpha didn't vanish completely. AMD licensed the 21264 EV6 FSB and it became the basis for the FSB of the Athlon XP (Socket A) and still lives on in the Geode NX. And of course the SMT implementation in the Pentium4, and now the Core i7, comes from the 21464 EV8. – Jörg W Mittag Jul 10 '09 at 23:44
  • @Jörg W Mittag Socket A first existed as Slot A, which of course, was the genesis of that bus, but yeah, excellent reference! – Avery Payne Jul 11 '09 at 20:39

Seems like other than some interesting history on the development of NT, we're all pretty happy to dump whatever version of Windows we were using and purge it entirely from our memories :)

That, and there's not much legendary or esoteric about Windows!

  • Re: see Jörg W Mittag's answer and the comment below it, there's LOTS of legends surrounding windows, some true, some not. No down-vote from me (not my style) but you might want to research this a bit further...there's plenty of hidden lore out there. – Avery Payne Jul 11 '09 at 20:41

Not the answer you're looking for? Browse other questions tagged or ask your own question.