Talk:Windows NT/Archive 1

Re Naming

It has been suggested that Dave Cutler intended the initialism 'WNT' as a play on VMS, incrementing each letter by one.

That's an explicit reference to PF Zachary's book and PF makes no such allegation. Brett Alexander Hunter (talk) 07:41, 23 October 2021 (UTC)

Ubuntu Vandalism

I'm new to Wikipedia so I couldn't work out how to change this. But in the infobox on the left-hand side, under the "Latest stable release", there is some boldface text proclaiming that Windows is "bad", and that Ubuntu is far better. Whilst I may or may not agree with the sentiment, this looks like vandalism to me. Qlexander (talk) 10:31, 23 April 2013 (UTC)

Yes, that vandalism was a bit complicated. Somebody vandalized the template Template:Latest stable software release/Windows NT. The infobox is generate from the template Template:Infobox OS, and that uses conditional tests to figure out whether to include the OS's "Latest stable software release" template. Somebody else undid the vandalism to Template:Latest stable software release/Windows NT. Guy Harris (talk) 17:59, 23 April 2013 (UTC)

Comments

I don't know the origins of NT well enough to write even a good stub article. I simply neutralised the comments that existed before, and took out the BSoD reference entirely --Colin dellow

The first sentence is not even English. This article, from the point of view of the informed user needing more information, is largely useless. Needs to be edited to bring out the facts (release dates, versions, features, which parts of the technology have been carried into Win2K and XP) and remove most of the opinions, which help no one other than the writer. --BK

Microsoft vs IBM

There are a number of mistakes regarding the history of MS-Windows in this and related articles. First up NT, was NOT based on OS/2 v.3 which was released a few years after NT. Nor was OS/2 writen by Microsoft, it was a joint IBM/Microsoft project to produce a mature OS for the home PC market, just do a search for a famous video clip of Bill Gates in 1989 cooing over how OS/2 is going to be the OS of the 1990's. It was only after he saw the sales figures for a semi workable GUI environment that he realised selling vapour ware would be more profitable than a joint mature & working product which he'd have to share with IBM, who's engineering skills well deserved. To this end Mr Gates had his people do a universal text-processor change of the source code for OS/2 ver. 1.2 (it may have been 1.3, but I think it was 1.2) replacing the tag 'DOS' with 'WIN'; recompiled, and called that Windows-NT version 1.0 . This was the reason that NT at first 'could' run OS/2 programs, because they couldn't stop it from doing so for the first half dozen releases. You must remember OS/2 1.x was a 16 bit OS, IBM re-wrote from the ground up as a pure 32 bit OS for ver. 2.0 which had the Windows 3.0 'emulator'; which was easy as 90% of the Windows code was written by IBM code cutters under contract & that contract which gave IBM access rights was the reason that IBM also wrote bug-fixes for MS-Windows 3.0 & 3.1; however, Microsoft took IBM to court to prevent IBM from releasing the fixes & claiming IBM had no rights to any of the Windows code. The court case was bogus but MS kept in in the courts from 1992-1995 when it of course became meaningless.

Basicaly, Windows NT is the Microsoft version of OS/2.

MS-Windows95 on the other hand was a Desktop system mimicing the OS/2 Desktop system - complete with weaknesses as well as strengths- sitting on top of MS-Dos v.7 The modern Microsoft Windows is a Microsoft version of the OS/2 Desktop manager on top of a Microsoft version of OS/2. And OS/2 was a fine OS for its time, and for its market; but it nor anything based from it is NOT suitable for todays network environment.

I see no reason for editing the foregoing other than to fix the mis-spellings. The article is substantially correct in its analysis of Microsoft's game-plan. sjc

I'm no MS fan either, but I don't like the sneering tone in which this article is written. The analysis is indeed correct, but could be described in a somewhat more neutral way

IBM didn't "produce" OS/2; Microsoft did. It was released as Microsoft OS/2 1.0, a complete operating system, entirely produced by Microsoft. IBM bought it, improved upon it, and released it under their name later. My impression at the time was that MS used it as something of a "testbed" for how to write an OS, then pawned it off to IBM and started over with NT. Internally, most of the MS team who produced OS/2 were the same folks who produced NT (For example, the OS/2 component test team directly became the NT component test team, which was where I spent my brief employment at MS). --LDC

NT vs OS/2

As I recall it, it was a joint venture between the two of them. I'm sure each side would claim it did the lion's share. I recall that significant amounts of IBM code went into OS/2, including in particular the graphics system. There was plenty of disagreement, because IBM wanted it to fit into their SAA architecture, while Microsoft wanted to keep compatibility with the Win16 GDI. In any case, I know specifically that OS/2 versions 1.3 and 2.0 were released under both Microsoft and IBM brands. That simply could not have happened if IBM just bought it later. The schism happened after OS/2 2.0; IBM went on to release OS/2 3.0 and Microsoft released NT 3.0. That's why there was no NT 1.0 or 2.0

--Alan Millar

I'm sure you're right about IBM being more involved; I used Microsoft OS/2 1.0, but I didn't join MS until after the team had moved to NT, so I can't speak to the details in between. I can, however, state with some confidence that the reason the first release of NT was called "3.0" was to synchronize its version number with that of the currently-popular Windows 3.0. By the way, if you have an old copy of NT 3.1 lying around, you'll find the email addresses of the team, including "leecr", in the easter egg. --LDC

What do you mean by 'release of NT called 3.0'? AFAIK there wasn't any release of such OS, maybe something deeply internal and not seen by anyone out of MS. --tyomitch 21:36, 1 September 2005 (UTC)

NT3.1 was the first public release. I always believe that one reason for using the 3.1 term was the same reason that Windows for Workgroups took the number 3.11: to retain the the Netware client licenses, which were restricted to windows 3.1. I'd be hard pressed to find evidence for that though. SteveLoughran 21:28, 9 January 2007 (UTC)

Agreed. Ive actually got a copy of Nt3.1. It was superceded by 3.5 and 3.51. It uses program manager like windows 3.1 but clearly says Windows NT on it. Its got networking, and is a fully 32 bit os. It can still run some modern programs, in fact.134.36.92.18 (talk) 00:06, 26 April 2009 (UTC)

Although I don't have it direct from Cutler, I hardly knew him and he knew me even less, I am as certain as I can be that the VMS => WNT ploy was intentional. Cutler was fond of puns and word play and worked a number of jokes into RSX, where I worked with him. See Talk:RSX-11 for some examples that are verified out of my own experience and some indications as to his personality. He was a profane and sarcastic individual who felt unappreciated by DEC and was glad to go to Microsoft. It would be absolutely like him to rub it in. So no, the VMS=> WNT thing isn't verified, but I know it happened. Ortolan88 20:47 Aug 4, 2002 (PDT)

Discontinued software?

This article was added to Category:Discontinued Microsoft software this morning. However, is that really an accurate classification for NT?
On the one hand, Microsoft has not marketed new products in this line of software under the trademark "NT" since the release of Windows 2000. On the other hand, each release of Windows since that time has reported itself as, inter alia, "Windows NT 5.0" (in Windows 2000), "Windows NT 5.1" (in Windows XP) or the like, which suggests that Microsoft continues to use the appelation "NT" at least internally.
—Ryanaxp 14:02, Jun 14, 2005 (UTC)

In lots of internal things, microsoft products still use NT. E.g lots of active directory binaries use the acronym NTDS (NT Directory Service) 134.36.92.18 (talk) 00:07, 26 April 2009 (UTC)

It should be undiscontinued. MS frequently refer to the current windows kernel architechture as the NT architechture in developer docs —Preceding unsigned comment added by 134.36.93.46 (talk) 03:03, 21 May 2009 (UTC)

Missing version?

NT also has several different flavours not mentioned here, I think. I know that Terminal (Server) Edition also exists for NT 4.0. Advanced Server and Server exist in Windows 2000. Windows 2003 has at least Enterprise Edition. Jdstroy 03:07, 2004 Aug 13 (UTC)

Why no mention of NT 4.5? Edit: for those who claim there's no such thing, Windows NT 4.5 Small Business Server and Windows NT 4.5 BackOffice Server.

OK, thats what I did, but you better comeback with some screenshots. :) --Noypi380 15:12, 9 August 2005 (UTC)

Ars-Technica says that NT 4.5 is a designation for NT4 SP4. Searching Google Groups shows that this is indeed a widely-used name. They also say that build number is 1433. So what, should we mention this version in the article? --tyomitch 21:36, 1 September 2005 (UTC)

I believe it's called "Windows NT Small Business Server 4.5" or "Windows Small Business Server 4.5", not "Windows NT 4.5". There is no Windows NT 4.5. AlistairMcMillan 22:40, 1 September 2005 (UTC)

Isn't "Windows NT Small Business Server 4.5" considered Windows NT? If it does, then it belongs to this article. --tyomitch 22:53, 1 September 2005 (UTC)

Small Business Server is just a version of Windows that is sold as an integrated package with a bunch of other crap (Exchange etc). It isn't a new version of Windows. AlistairMcMillan 23:06, 1 September 2005 (UTC)

OK, but History of Microsoft Windows#Windows Server 2003 lists SBS as an edition of Windows Server 2003, together with the 4 editions mentioned in this article. 1) shouldn't the NT 4.5 SBS, be it an edition of NT4, be added to the list of NT4 editions? 2) shouldn't the edition lists of the two articles be merged, probably at History of Microsoft Windows, where I think it belongs more? 3) why then NT 4.5 SBS has a build number so strange, not matching any other NT release? tyomitch 23:44, 1 September 2005 (UTC)

I now see at this Microsoft page that SBS 4.5 (that's what it's called) is in fact NT4 SP4 + various stuff. Now what is this version to be considered? An edition of NT4 or separate Microsoft product unrelated to NT? SBS doesn't have it's own article, so it has to reside somewhere in the Windows articles. Having build number 1433 is most likely a hoax. --tyomitch 00:11, 2 September 2005 (UTC)

What part am I not explaining. They take a pre-existing version of Windows Server (originally Windows NT 4.0 Server, then Windows 2000 Server and now Windows Server 2003), then add on a bunch of programs (Exchange Server, Proxy Server, etc), stick them all in a box and sell them. It is not a separate version of Windows. Small Business Server is a separate product, not a version of Windows. All it has just now is a redirect Microsoft Small Business Server, please feel free to expand on that if you want. AlistairMcMillan 00:37, 2 September 2005 (UTC)

Thanks for clarifying that. And what about redundant edition lists? (Again, SBS 2003 is listed as a Windows edition at History of Microsoft Windows –- should it be corrected there?) --tyomitch 01:05, 2 September 2005 (UTC)

Damned if I know. That page is a mess. We are listing every single version of XP, even if the only difference is the inclusion of a single bundled app, but we don't list all the available versions of NT. AlistairMcMillan 01:34, 2 September 2005 (UTC)

Windows 2000 has a datacenter variant. There is NT embedded, windows 2000 advanced server, nt4 terminal server edition (joint program with citrix i think). And BackOffice server "NT 4.5" is just NT4 with exchange and mssql etc... bundled.134.36.92.18 (talk) 00:09, 26 April 2009 (UTC)

Microkernel design?

Does anyone have a source for this...

No, not now, but have been acquainted with people who worked with the Tribe in Redmond. Yes the idea was microkernel. But Billg waited two full years before using proxy to inform Dave that he wanted a workstation version too, a version with a GUI, and Dave really hates GUIs. He was given people from Win16 to help with something he didn't even want. As Win16 (Win3.1, W4W, Win95) got faster, Dave decided to forego microkernel to keep his speed advantage, moving the user module into kernel mode, something that resulted in more BSODs under NT 4.0. Brett Alexander Hunter (talk) 07:51, 23 October 2021 (UTC)

Originally, they tried a microkernel design, but failed to create a working version that used a microkernel. They ended up going with a monolithic kernel that integrated more functions into the kernel.

This was added by User:216.119.235.214 over a month ago. [1] I'd love to know if this is true. AlistairMcMillan 22:23, 30 August 2005 (UTC)

Someone recently deleted the mention of Windows NT as an example operating system on the Mach Kernel page. I've seen no evidence that NT uses Mach, and at best only have seen comparisons and hints at inspiration from Mach. I'm deleting the mention of Mach from the "See also" section. --64.222.109.153 22:17, 19 March 2006 (UTC)

At least Microsofts own documents tells that NT operating system use microkernel. But because the NT operating system is not developed as pure microkernel structure, where the microkernel alone runs in kernel space, but some of the OS services what are running typically in userspace when making OS with microkernel structure, has moved to kernel space with microkernel. The article has it wrong that graphical subsystems etc were integrated to kernel, because they were moved to kernel space but not integrated to microkernel. The Windows NT OS is the hybrid kernel, but there is no actually such structure, just a marketing. Everyone can ask that if Windows NT is hybrid kernel, why does hybrid kernel has a microkernel? Sounds very stupid to do a technical structure to have kernel inside of the kernel. Truth is, that the microkernel is inside of the operating system, and the operating system is the so called "hybrid kernel". Just like the monolith kernel is the operating system and microkernel structure idea is to slice the monolith OS for pure microkernel and sliced OS parts to userland as OS services or OS servers running in supervisor mode together. And the difference of monolith and "hybrid kernel" is that monolith still is alone an OS, but the hybrid still has some of the OS parts moved from kernelspace to userspace and microkernel is still loosely from other OS parts in kernelspace. I dont believe that Windows NT microkernel is actually a Mach kernel, even that Mach designer worked for Microsoft NT project, so citation for that would be needed, but I dont say it is impossible. I think that Mach mention is good because the NT operating system uses the microkernel and Mach is good example of such kernel. The article itself needs a mention that Windows NT OS use the microkernel because the OS is not just the hybrid kernel. Golftheman (talk) 10:08, 21 November 2008 (UTC)

Some of "Microsoft's own documents" have used the word "microkernel" (some versions of the Resource Kit docs, for example) but they in turn did not use the word "kernel", implying that what they were calling the "microkernel" is simply the "kernel". David Cutler is on record as saying "this is not a microkernel design" (1992 NT Driver Developer's Conference), and the most authoritative MS reference on the OS design, Windows Internals by Russinovich and Solomon, explicitly states the same (see page 36). And no, NT is not based on the Mach kernel nor on a Mach-like design. Rather it shows a clear line of evolution back to VMS; anyone familiar with the internals of both will readily confirm this. As for your claim that "the operating system is the so called hybrid kernel", no - there is a lot of the operating system that is outside the kernel - things like the executive, device drivers, HAL, the Win32 API, the graphics subsystem, Explorer (the primary UI), etc., etc. Such a claim isn't compatible with reasonable definitions for both "operating system" and "kernel". Jeh (talk) 21:41, 21 November 2008 (UTC)

Windows XP Professional x64 Edition is what version of NT

Windows XP Professional x64 Edition is branded as Windows XP (NT 5.1), but based on Windows Server 2003 (NT 5.2). Which NT version number does Windows XP Professional x64 Edition claim to be? --Pmsyyz 23:33, 17 September 2005 (UTC)

A person running it reported that it's 5.2.3790.2180 (fixed the article now)--tyomitch 23:54, 17 September 2005 (UTC)

Yeah its 5.2. Also the most recent service pack for it is SP2 (uses its owne modified server 2003 serivoce packs)134.36.92.18 (talk) 00:10, 26 April 2009 (UTC)

Longhorn/Vista

http://search.microsoft.com/search/results.aspx?st=b&na=88&View=en-us&qu=longhorn Microsoft sites say Vista has replaced longhorn.71.28.243.246 19:58, 13 October 2005 (UTC)

Does the article say otherwise? If not, what's the point of your comment? --tyomitch 21:22, 13 October 2005 (UTC)

Vista has not been released yet (as of 21 February 2006) and is expected around the end of 2006 (although it could slip into 2007 - my opinion). The article has now been reverted the to reflect this again. Microsoft has info about the Vista beta [2]Imagine_B 10:43, 21 February 2006 (UTC)

Disk, not Disc

I reverted an anonymous edit that changed disk to disc. Looking at the wiktionary definition of disk, disc is the North American variant of disk; using this description it is best to keep the terminology consistent. There are roughly three occurances of disk so changing one of them to disc is not a good idea.

The other way to look at it, is in international English (according to wiktionary definition of disc) disc implies "optical disk", in which case hard drive space does not qualify to use disc. BigNate37 19:58, 26 June 2006 (UTC)

Windows NT / The Single UNIX Specification / POSIX

Is it necessary to make the comparisons between UNIX and Windows NT? I realize that there is a great schism between the masses of Windows NT and UNIX users, but I don't feel that it would be appropriate to make the claims of:

• "features comparable to UNIX"
• "generally not compatible with UNIX in terms of programming APIs"

I'd like to point out that, up to Windows NT 5.0, there *is* a POSIX subsystem (posix.exe and pax.exe) deployed on a standard installation; thus, Windows NT 5.0 and earlier had built-in support for a UNIX platform. This doesn't really fit the description of "features comparable to UNIX," as it contains a UNIX platform. Windows NT's "subsystem" architecture was designed in a manner similar to the idea of "universes" on a UNIX system--it allowed a user to run programs in multiple platforms, including the Win32 subsystem (and 16-bit DOS/Win16 subsystem), the POSIX subsystem, and the OS2 subsystem.

In later versions of Windows NT (5.1, 5.2), the support for the POSIX subsystem was dropped; however, an add-on package from Interix (and now Microsoft) called "Windows Services for Unix" allows for the use of the POSIX subsystem on Windows NT 5.1 and Windows NT 5.2.

Although Windows NT 5.0 and earlier had support for the POSIX subsystem in a standard installation, it lacked the tools that went along with a standard UNIX install. (It's like having a Linux kernel, but no GNU!) The rest of the runtime environment was left in the Windows Services for Unix package.

I believe the second quote (above) is incorrect. SFU is "a POSIX.1-conforming system" [3]; thus, it has source compatibility (up to the POSIX.1 specification, see POSIX) with "UNIX."

Thoughts? Jdstroy 04:36, 4 July 2006 (UTC)

Windows Posix != SFU. The Posix API was a peer subsystem for NT, alongside the OS/2 subsytem. It used features that are not exposed to the windows API, but were clearly in the NT kernel

• Case sensitive filenames. You could even create two files of the same name and different case in a directory.
• hierarchical process trees. When you kill a process, you kill its children.
• fork(). There is no direct equivalent in the Win32 API.

Its purpose was to have a complete implementation of Posix that was absolutely no use to code against, which was achieved by not offering any network API, no graphics API, or anything other than a command line. Why do it? So people purchasing NT boxes on government deals could check off the "Posix compliant" box, which was often a requirement.

Services for Unix are services that let windows integrate with linux, and vice-versa. It includes things like an NFS client and server. It is not a unix or posix API. The closest there is the cygwin API, which runs on top of Win32, bridging over to that API to access everything posix left out. It does, however, lack access to those features that are only done properly in the real Posix subsystem, like fork().

SteveLoughran 21:24, 9 January 2007 (UTC)

SfU is not the old "Windows POSIX" but it does provide a separate NT subsystem implementing the POSIX API in addition to the aformentioned NFS client and server -- I'm not sure if the NFS stuff is built upon that API, but the shell and compiler (yes, MS ships gcc!) in SfU certainly are. (Yes, I realize I'm responding to comments that are two and three years old. Can't... let it... go....) — jhf (talk) 09:30, 17 February 2009 (UTC)

It may be ineresting to note that MS bundle a number of unix utilities with windows. In the windows release notes its got the disclaimer (the one BSD type licences require you to have). I read that utilities like FTP were rebuilt by MS from sources from the OpenBSD project. —Preceding unsigned comment added by 134.36.92.18 (talk) 23:47, 25 April 2009 (UTC)

Huh?

Since increases in processor performance, Windows XP and Windows Server 2003 are actually the latest versions of Windows NT, though they are not branded as such for marketing purposes.

How does the fact that Windows XP and Windows Server 2003 are not branded as NT follow from the fact that CPUs are now faster?

MSTCrow 02:07, 8 July 2006 (UTC)

changed. SchmuckyTheCat 02:20, 8 July 2006 (UTC)

NT on Alpha

"Released versions of NT for Alpha were 32-bit only"

This doesn't really make sense and I don't think it is even possible. The Alpha is purely a 64-bit architecture. --Afed 16:55, 18 August 2006 (UTC)

NT on Alpha supported the same 32-bit Win32 API and 32-bit addressing as the x86, MIPS and PowerPC ports of NT; although early 64-bit NT development was done on the Alpha platform, this was never productized. See Compaq's Alpha NT EOL letter which explicitly mentions 32- and 64-bit NT. OpenVMS Alpha was also 32-bit until v7.0. Letdorf 23:48, 18 August 2006 (UTC).

Indeed. The shipped versions of NT for Alpha were pure 32-bit operating systems and supported only 32-bit applications. Same for the first versions of VMS for Alpha, for that matter.

It's true that the chip has no 32-bit mode. This is handled through what amounts to sign extension: Any memory reference on an Alpha has to be done by loading the pointer into a register. And any code compiled for 32-bit Alpha, when loading a pointer into a register, treats it as a signed number, copying bit 31 of the pointer into bits 32-63! Thus the 32-bit OSs on the Alpha populate the very first 2 GiB and the very last 2 Gib of the 64-bit address space, in a manner very similar to the 48-bit "canonical" virtual addresses on X64.

However, unlike 32-bit mode on e.g. X64, all 64 bits of the Alpha's GPRs (and all of the GPRs, for that matter) were available for "large integer" calculations in this "32 bit" environment!

MS released 32-bit NT for Alpha through NT version 4, and shipped some 32-bit Alpha beta versions of Win2K. This continued after CPQ dropped support on NT for Alpha, but with no intention on MS's part of shipping the code. It was done for two reasons: 1) There was Alpha hardware but no Itanium hardware; and 2) they wanted to maintain a code base for more than one processor architecture, a practice that made the (much later) X64 port much easier. But no 64-bit NT code was ever shipped, even in beta form. (I assure you that if it had been, I'd be running a copy!) I imagine that now that they have three architectures to support the Alpha branches have not seen further development. (the above shamelessly plagiarized from my own text on the talk page of Architecture of Windows NT ) Jeh (talk) 03:23, 14 January 2008 (UTC)

What does "to complement workstation versions of Windows" mean?

The article says:

It was originally designed to be a powerful high-level language-based processor-independent multiprocessing multiuser operating system with features comparable to Unix to complement workstation versions of Windows that were based on MS-DOS until 2001.

What does "to complement workstation versions of Windows that were based on MS-DOS until 2001" mean here? I'm not sure I'd call the DOS-based versions of Windows "workstation versions" - NT was better suited to "workstations" than the DOS-based versions. (The workstation article speaks of UNIX, but that's arguably an error; machines used as scientific and engineering workstations often run NT.)

I'd be inclined to say that NT was originally mainly a workstation and server version of Windows, complementing the DOS-based lower-end consumer and business desktop versions of Windows. It's now replaced the DOS-based versions in the latter market. Guy Harris 00:05, 19 August 2006 (UTC)

NT confusion

Aha, so 'NT' is expanded to 'New Technology' for marketing purposes. And also for marketing purposes Windows 2000 had the slogan 'Built on NT technology.' So this expands to 'Built on New Technology technology.' This must be a joke, right? iNic 02:40, 23 November 2006 (UTC)

Yeah, I was thinking the same thing while I was typing in my PIN number at the ATM machine a few days ago.

(You're not expecting something done for marketing purposes to make logical sense, are you?) Guy Harris 03:25, 23 November 2006 (UTC)

I do expect marketing to make logical sense. If the marketing doesn't make logical sense why would their products make logical sense? For example, marketing for New Age products doesn't make logical sense—but neither do their products. So that is fine with me. But maybe MS products doesn't make logical sense either? In that case I will rest my case. (BTW, I've never heard anything but PIN code, never PIN number.) iNic 02:25, 25 November 2006 (UTC)

If the marketing doesn't make logical sense why would their products make logical sense? Becuase they're two totally different things, perhaps? What a marketer has to say about a product has absolutely no bearing on the real, actual product itself. The job of a marketer is to do market research, find out what the "hot issues" are, what the latest techniques in wording, format, and style are, and apply those things to their work. -/- Warren 04:42, 25 November 2006 (UTC)

NT is not about New Technology, but about a HAL9000 like joke... If you take Windows NT, remove the "indows" you will have WNT, and if you do a ROT-1 you will get "VMS" (it's much like "HAL" in "2001: A Space Odyssey", if you to a ROT1 you will get "IBM") - Godzil --212.157.49.189 14:45, 26 September 2007 (UTC)

The NT repetitive thing is outlined in this article for any of you out there who want to read about it. Emprovision (talk) 19:12, 27 May 2008 (UTC)

processor-independent or not?

Like Unix, NT was written in C, a high level language. It can be recompiled to run on other processor systems, at the expense of larger and slower code.

Two problems with this statement: first, what's the basis of the comparison? "Other", "larger", "slower" compared to... Intel x86? Intel i860? MIPS R4000?

More important: A priori there's no reason to believe the claim that recompiling on another plaform gives larger and slower code compared to the original. Without a reference and clarification, this statement should be removed. Bhudson 21:02, 6 January 2007 (UTC)

I suspect that sentence was supposed to mean that code written in C and compiled to a particular processor's machine code is larger and slower than code written natively in assembler language for that processor. Part of the problem is the "it can be recompiled to run on other processor systems" clause; it would probably be better stated as "it can be compiled to run on more than one type of processor", or something such as that. In addition, the two clauses should probably be separated; it's not the recompiling that adds the expense, it's writing it in a higher-level language (although a C compiler can probably outperform assembler-language programmers with lesser skills).

In addition, NT's greater code size, and longer code paths, are probably more due to NT being more offering a lot more capabilities than traditional Windows than due to it being written in C. Guy Harris 23:59, 6 January 2007 (UTC)

C is slower than assembly, that was used before for large parts of win 3.1 and dos. When porting NT to ALPHA it was boasted (sorry, no ref) that a large of the OS just ran after a compile on the new platform. THis flexibilty would have never beeen possible with assembly, but C is slower than assembly (at the time of the release of window NT 3.5 this was even more true than today) :Leuk he 12:35, 26 June 2007 (UTC)

Just wondering (I could be wrong), but wasn't NT written in C++? Emprovision (talk) 15:46, 27 May 2008 (UTC)

-If i remember rightly, when NT war originally made (nt 3.x) the dev machines used were deliberately not intel x86 machines. The reason behind this was to stop the developers using nonportable hacks. —Preceding unsigned comment added by 134.36.92.18 (talk) 23:51, 25 April 2009 (UTC)

WinME came out before NT5

Alistair: WinME was a bastard operating system that was released in mid-1999, because MS could not get NT5 out the door in time for the Christmas 1999 season. NT5, rechristened Windows 2000 at the last minute by the marketroids, was released in a big coast-to-coast bash featuring Carlos Santana in mid-February 2000.

By the way, I was a beta tester for both the x86 and AlphaNT versions of Win2k... In fact, I still run RC2 (build 2128) on my DEC AlphaPC 164SX. Dan Schwartz Discpad 23:22, 13 March 2007 (UTC)

That's funny. Why did Microsoft announce the release in September 2000, if it came out in 1999? http://www.microsoft.com/Presspass/press/2000/sept00/availabilitypr.mspx AlistairMcMillan 23:24, 13 March 2007 (UTC)

The press release is correct. I remember the news spots about 2000 not being for home users, etc, etc, and then the big hullabaloo a few months later about the "long anticipated" (and highly disappointing) release of ME. -amp_man 26 June 2007

Yeah MS wanted to combine the home (DOS) and business (NT) lines into one (like they did eventually with xp) in tje late 90s, but Windows Neptune (codename of the planned OS) was not ready for home use just yet, so they released win2k and ME as a stopgap (me was made in like 6 months).

Thats why XP came out only like 9 months after windows 2000, and they are very similar. Windows XP was basically windows 2000 + the home stuff like the application compatibility wizard, network setup wizard etc.... planned for neptune.

If you try the windows neptune alpha (some on the net) you will see what i mean. Its basically a very buggy win2k, with some wizards that didnt emerge until xp (like the beginnings of a welcome screen implementation) —Preceding unsigned comment added by 134.36.92.18 (talk) 23:54, 25 April 2009 (UTC)

Windows Home Server.

The article says Home Server will be NT 5.2. If there is a source to this claim please post here if not I will change it to:

NT Ver.	Marketing Name	Editions	Release Date	RTM Build
Unknown	Windows Home Server	Unknown	2007 (expected)	Unknown

Chetblong 20:03, 1 June 2007 (UTC)

Windows Home Server is a specialized build of Windows Server 2003, which is where 5.2 comes from, but screenshots like this one point to it being 6.0. Hard to say one way or the other without something more reliable. -/- Warren 23:39, 1 June 2007 (UTC)

Yeah that does looks a lot like Windows Vista. I'm thinking it's going to be somthing in the NT 6.0+ range, but since there is no info I won't put that on the article. I'll wait 5 days and see if any source comes up if not I'll change it to Unknown.--Chetblong 23:49, 1 June 2007 (UTC)

It looks like Vista because that's a screenshot of the client software, not of the server itself. :) -/- Warren 07:18, 3 June 2007 (UTC)

I went ahead and changed it to Unknown. -- :) Chetblong 19:02, 7 June 2007 (UTC)

i imagine it will be listed as '6.0'.. vista is also based on server2003, so it should be 5.2 (or 5.3?) but i suppose for marketing/effect they made it 6.0, in which case, so will home server i'd guess.~blab

Windows Vista is 6.0 because of the major improvments over 5.2. Windows Home Server does not have these. Josh 15:46, 6 October 2007 (UTC)

It is 5.2. I own it. Its a version of 2003 Small Business Server with domain functionality and IIS removed, among other things + addons like the new volume manager and backup.

The file version it comes up as (on nlite etc...) is 5.2

fat is faster than NTFS

[all nt <vista can be installed to fat..] "at the expense of speed and security" no, fat is faster than NTFS for normal use on all but the very largest drives.~blab

 ;Only on small partitions running DOS! Per KB: "over 200 MB the FAT file system should not be used. This is because as the size of the volume increases, performance with FAT will quickly decrease." per http://support.microsoft.com/kb/100108 . Also consider problems like more than a few thousand entries in a directory (like Internet Cache).

"Limitations of FAT32 File System". Microsoft Help and Support. 2004-12-16. http://support.microsoft.com/kb/184006/en-us Chen, Raymond (2006). Microsoft TechNet: A Brief and Incomplete History of FAT32. TechNet Magazine July 2006 http://groups.google.com/group/microsoft.public.win98.fat32/browse_thread/thread/1fa5783ca7a26a5b

Shjacks45 (talk) 13:20, 13 April 2010 (UTC)

Hardware requirements

The Hardware requirements section doesn't mention year difrences between the respective operating systems, something I feel may make it be unbiased against Microsoft for the percieved software bloat (noted, they rely less on application optimization and more on Moore's Law in order to get speed increases). If the years were listed in the table, perhaps it would be more effective at showing the data? - 68.228.56.158 19:13, 20 July 2007 (UTC)

Stable version?

Which is the most stable NT version? I know the kernel is refined all along the way, and NT 5 (Windows 2000) included a major architecture redesign. However, what about the whole system's buglessness and stability? In Yahoo! answers one said it was Windows 2000 that was the most stable version. I myself felt that too (patched at least up to year 2004). In my experience XP is also stable, however with some strange user mode hang-up situations: when switching users, logging off, etc. --211.167.159.70 08:06, 14 August 2007 (UTC)

-the server builds are the most stable IMHO. Server 2000 or 2003 have never crashed for me under general use, unlike XP and 2000 Workstation —Preceding unsigned comment added by 86.16.160.17 (talk) 21:45, 2 January 2008 (UTC)

Depends on what you consider stable. Downtime? In which case probably Server 2008 - its got self-healing NTFS and hotpatching. Windows 2000 is mature - but that doesnt nececerially mean stable. —Preceding unsigned comment added by 134.36.92.18 (talk) 23:58, 25 April 2009 (UTC)

Versioning

Does the controversy concerning the minute different versions of NT Server (4.0) and NT Workstation (4.0) deserve to be brought up in this article? http://www.oreilly.com/news/differences_nt.html —Preceding unsigned comment added by 75.74.240.92 (talk) 02:56, 27 November 2007 (UTC)

Releases

Release list should include Home Server and processor type for all releases. Helpsloose 21:08, 20 December 2007 (UTC)

C vs. ASM bullocs

NT was written in C, a mid-level language. This means that it can be compiled to run on several processor systems; however, the code produced by the compiler is larger and slower than assembler code written for a particular processor[citation needed]. For this reason, NT was not favored initially for use with slower processors with less memory.

As a CS engineering student, I can almost certainly say this is wrong. Code produced by a compiler is not inherently slower, and it will CERTAINLY not be the reason why a certain OS cannot be used for a certain machine. Heck, Linux avoids ASM code where slightly possible, only very lowlevel function that cannot be described in C will be programmed in ASM. —Preceding unsigned comment added by 193.190.253.149 (talk) 15:52, 11 January 2008 (UTC)

What it means is that NT is slower than 9x.

9x had lots and lots of asm code, which was highly optimesed for x86. —Preceding unsigned comment added by 134.36.92.18 (talk) 23:59, 25 April 2009 (UTC)

Remember kiddies: Well written C code is faster then badly written assembly. What language the OS was written in is NOT enough to indicate the speed it runs at. Nevermind all the diffrences in CPU/GPU/RAM/HDD requirements/designs over the years. Then again, how are we measuring "speed" of an OS anyway? --151.190.254.108 (talk) 20:12, 24 June 2010 (UTC)

XP 64 bit editions

What do you think of the table now? I confess I'd forgotten completely about XP 64-bit Ver 2003! (and MS did too, very soon after...) I have to run out now, but later I'll put in a link to that press release.

yeah, there is XP x64 (Intel Itanium architechtute) and XP Pro x64 (the common x86_64)

XP x64 (itanium) was pure xp SP1. XP Pro x64 was based on Server 2003 x64

Microsoft has never referred to Itanium as "x64" - this is what they (and Sun) call AMD64/EM64T/Intel64/x86-64. The Itanium version of XP was called Windows XP 64-Bit Edition. Letdorf (talk) 13:19, 27 April 2009 (UTC).

The table, with processor architectures added

Do you think the supported processor architectures should perhaps have a column of their own? This might mean a few more new rows, but it's arguably useful to have it in the table. Jeh (talk) 03:15, 14 January 2008 (UTC)

Sounds good. - Josh (talk | contribs) 03:21, 14 January 2008 (UTC)

How's this? /Sandbox/Windows NT processor architecture table Jeh (talk) 04:46, 30 January 2008 (UTC)

Looks good. Letdorf (talk) 10:45, 31 January 2008 (UTC).

Looks good, although I don't see the point in specifying the service pack. - Josh (talk | contribs) 14:39, 7 February 2008 (UTC)

In those cases the indicated arch was dependent on the SP. In particular, x64 on Server 2k3 only appeared with 2K3 SP1. XP's x64 was actually 2k3 SP1 also. Jeh (talk) 15:25, 7 February 2008 (UTC)

XP 64 bit (Intel Itanium) was based on pure xp, not 2003. It was only supported up to SP1 i think. XP ~Pro~ x64 is supported up to SP2 and is based on Server 2003. Dont get them confused.

134.36.92.18 (talk) 00:12, 26 April 2009 (UTC)

Windows 7 kernel version

Do we even have a source for this? If not, it should be removed. I doubt Windows 7 will be NT 6.1, especially considering it's Windows 7, which comes after Windows NT 6/Vista. 99.149.127.80 (talk) 06:43, 5 April 2008 (UTC)

The one Windows 7 build outside Microsoft has the version number 6.1. - Josh (talk | contribs) 07:11, 5 April 2008 (UTC)

The final version number is not known; 6.1 is just the current kernel version. — Wenli ^{(reply here)} 23:49, 29 May 2008 (UTC)

MS said (in a webcast on msdn discussing windows driver development for windows 7 for OEMs) that windows 7 will be NT 6.1, as will windows server 2008 R2. This is in order to keep compatibility apparently. —Preceding unsigned comment added by 134.36.92.18 (talk) 00:01, 26 April 2009 (UTC)

I doubt it also that it will be NT 6.1, it's not logical, and further more my I heard that Microsoft named 'Windows 7' Windows 7 because they want to go back to the naming of their OS with the version number... Like they did in the beginning... Windows NT 3.1, Windows NT 3.5, Windows NT 3.51, etc... Wouser (talk) 13:08, 18 May 2009 (UTC)

Since when did Microsoft have to be logical about these things? Windows 7 RC has a version number of 6.1.7100[4] and it seems unlikely that they will change the major version number this late in the development cycle. Letdorf (talk) 13:21, 18 May 2009 (UTC).

It is DEFINATELY 6.1 thats what it says in the windows 7 release candidate and what i read in some prelease material aimed at OEM driver developers. So will be Server 2008 R2 (which will be based on windows 7, as opposed to the original server 2008 which was vista sp1). Windows '7' is purely a marketing term. it is NOT nt 7. —Preceding unsigned comment added by 134.36.93.46 (talk) 03:07, 21 May 2009 (UTC)

There's precedent for the official marketed version number and the actual OS-reported version number being different: Windows for Workgroups 3.11 actually reports its version as "3.10". --20:47, 4 May 2011 (UTC) —Preceding unsigned comment added by 69.116.117.48 (talk)

Windows Server developer Preview uses 6.2 Kernel Winver. — Preceding unsigned comment added by 79.237.254.50 (talk) 10:56, 8 February 2012 (UTC)

Version numbers serve several purposes:

• error reporting, so that, for example, a bug report can say that something crashed on version XXX of the OS - that might be useful information to the developer;
• decorative, so that a system can announce itself as being version XXX;
• letting applications know what they're running on.

Unfortunately, those can collide; if Windows 7 were to provide a version number starting with 7, some apps might freak out with ZOMG I DON'T KNOW ANYTHING ABOUT VERSION 7 OF WINDOWS!!!!!1111ONE!!!!!!!!!! rather than, say, assuming version 7 was just like version 6 (gee, ya think an OS vendor might want to preserve some level of backwards compatibility for all the existing application binaries out there?) until a problem is discovered. That's why Windows 7 supplies an OS version number (I don't think that's the "kernel" version number, I think it's a build number for the entire OS) of 6.x rather than 7.x, as indicated in the Windows team blog (see also The Old New Thing's explanation of why Windows 95's GetVersion function returned 3.95 instead of 4.0 for another version numbering headache). Guy Harris (talk) 19:44, 8 February 2012 (UTC)

Xbox kernel

I think this source is certainly more reliable than this one. Accordingly, if no one has any objections, the Xbox line should be removed. Or at least reworded so that it says the Xbox OS is derived from the Win32 APIs but built from the ground up. - xpclient ^Talk 11:41, 22 May 2008 (UTC)

I agree completely. It's much better to trust Microsoft (who made the Xbox) rather than another source. I've boldly changed it. — Wenli ^{(reply here)} 23:45, 29 May 2008 (UTC)

From Linux Journal: "Its 256KB ROM image contains a statically linked, stripped-down, Windows 2000-based kernel" Actually Microsoft has monetary interests in misdirecting potential "hackers," and the XBox developer blogger has a professional reputation at stake. They might be protecting Intellectual Property. OTOH, The reporter at linux journal has nothing to gain from being wrong or certainly not deliberately misleading readers. I'm in favor of letting the edit stand without more evidence, but you're reasoning and justification for the edit is misguided. Tumacama (talk) 00:49, 14 August 2009 (UTC)

Windows 7 info in the "Windows NT Releases" table"

Windows 7 is still a codename and not a "marketing name", Windows 7 is schedueld to be released in H1 2010, not H2 2009. —Preceding unsigned comment added by 129.241.150.119 (talk) 15:55, 30 August 2008 (UTC)

Duly noted in article. Letdorf (talk) 15:21, 1 September 2008 (UTC).

Windows 7 WILL be sold as "Windows 7". Its the release/marketing name now, not just a codename. —Preceding unsigned comment added by 134.36.93.46 (talk) 07:13, 22 May 2009 (UTC)

Windows NT originally stood for...

See this. I don't feel like creating an account to fix this/cite that part of the article. Check the second paragraph. —Preceding unsigned comment added by 69.84.252.227 (talk) 00:28, 31 January 2009 (UTC)

But this article already gives the N10/NT explanation and cites that URL. What do you want to change? Letdorf (talk) 10:40, 31 January 2009 (UTC).

BSD TCP/IP

There's some disagreement regarding whether the TCP/IP implementation is derived or imitating BSD. However, Winsock is derived from BSD. Tedickey (talk) 14:06, 4 April 2010 (UTC)

RE: XBox

see xbox, dreamcast I worked at Microsoft in 1997 when NT 5.0 beta code was forked to Xbox. No Win32 or other subsystems. Windows Native Mode, Drivers, and DirectX; all kernel mode stuff. Not sure of Xbox 360, but there was PowerPC NT 5.0 code available at the time (except for Internet Explorer). Dreamcast was WinCE with DirectX, bare bones CE will fit in "250K" but not directx. Most modern BIOS is 128KB (to 512KB+) which includes chipset setup and EFI boot code then shrinks to 64KB after boot. Fact is most games load their operating system from the game CD or game cartridge, see Playstation "disk swap" section. There was a project there to "secure boot" PCs by putting MBR code and boot files into ROM (Flash) to prevent "viruses". NTLoader, NTDetect, boot.ini et al can fit into 250K. PicoXP is only 14MB. Shjacks45 (talk) 14:48, 13 April 2010 (UTC)

Inconsistent time line for i860 vs i386

"Initial development was for the Intel i860 RISC then MIPS until i386 became available.[5][6] ". According to the wikipedia articles about these processors the i860 was produced from 1989 and the i386 from 1985. The references 5 and 6 are mostly circular and are not useful. —Preceding unsigned comment added by L00KnS33 (talk • contribs) 07:48, 11 August 2010 (UTC)

Good point, I've reworded (and moved) the passage to hopefully clarify this. 386-based PCs were, of course available at the time; it was a deliberate decision to avoid them for the first couple of years of NT's development. Not sure what you mean by circular references here - those sources are books, at least one of which predates WP by several years. Letdorf (talk) 12:52, 16 August 2010 (UTC).

Windows 2000 and 64-bit

Windows 2000 64-bit was codenamed "Janus".

— Windows 2000's history

So how come it's not on Windows NT#64-bit? Also see Talk:Windows 2000#64-bit. -79.179.200.231 (talk) 19:34, 9 January 2011 (UTC)

Although the article doesn't give an entirely comprehensive history of NT platform support, it does mention Limited Editions of Windows 2000 Advanced Server and Datacenter Server, Windows XP 64-Bit, and Windows Server 2003 Enterprise and Datacenter support Intel's IA-64 processors. Regards, Letdorf (talk) 22:10, 10 January 2011 (UTC).

Please read. Even if they're related, this topic is about 64-bit (for the OS) and not about IA-64 (for the processor). -79.179.200.231 (talk) 14:36, 12 January 2011 (UTC)

Please udnerstand. The editions that run on IA64 are all 64-bit versions. Jeh (talk) 07:17, 14 October 2011 (UTC)

Yes, that is true; and they are all listed there. What exactly do you want? Fleet Command (talk) 12:02, 14 October 2011 (UTC)

Windows NT server

Here's an old Microsoft page about it:

• http://web.archive.org/web/19961026231042/http://www.microsoft.com/Infoserv/IISInfo.htm

WhisperToMe (talk) 01:58, 14 October 2011 (UTC)

No, that's a page about Internet Information Server, the Microsoft Web server. Guy Harris (talk) 19:12, 26 May 2013 (UTC)

Windows 8 marks the beginning of the WinRT era and the end of WinNT era

Steve Kleynhans said: "Windows 8 is more than a major upgrade to Windows - it's a technology shift. We don't see technology shifts too often; the only other one Microsoft's client OS has gone through was the move from DOS technology to Windows NT technology, which began in 1993 and took eight years, ending with the introduction of Windows XP in 2001,"

News link: http://www.ibtimes.com/articles/356436/20120626/windows-8-marks-beginning-winrt-winnt-gartner.htm — Preceding unsigned comment added by 77.121.210.102 (talk) 15:38, 3 July 2012 (UTC)

Well, he's wrong. Or at best he's spewing marketing fluff. WinRT is no more a fundamental change to the operating system than was .Net. In fact WinRT is very much analagous to .Net. It's another set of APIs, another way of writing programs, that's all. It is still layered on the Win32 API (now called the Windows API), which is still layered on the Windows Native API, and you can still run Win32 and .Net apps (not, apparently, on the ARM editions, but that is a licensing issue, not an omission). And the Windows API (csrss.exe and all that) still owns the desktop. The true "Windows 8 internal architecture" diagram doesn't look much different from the one for Windows 7, or even Windows 2000, except for the new WinRT API layer out in user mode. Jeh (talk) 16:05, 3 July 2012 (UTC)

Xbox One

This also uses an NT 6.x kernel. Should it be mentioned? PizzaMan (talk) 12:59, 26 May 2013 (UTC)

With citations, yes. (That's presumably the "Windows" OS, not the "Xbox" OS, as per what the Xbox One page says about the Xbox One having a hypervisor and two OSes running under it.) Guy Harris (talk) 19:10, 26 May 2013 (UTC)

Kernel Version and Windows 3.1

The Page is Currently Locked and Contains Information that is Uncited and I Know is incorrect. Windows 3.1 Didn't Run on the NT Kernel it ran on the DOS kernel and I know this for a fact. Someone Should Correct this I think it's more than a simple mistake.

Windows NT 3.1 most definitely ran on the NT kernel. Windows 3.1 didn't, and this page does not claim that it did. Guy Harris (talk) 22:18, 5 October 2013 (UTC)

microkernel approach was essentially a dishonest approach (sic)

The quote as given is not suitable for the technical description, since the point of it was to accuse competitors of being dishonest. There's no technical issue presented in the comment. TEDickey (talk) 13:16, 13 October 2013 (UTC)

So it should go into a criticism section? I mainly intended it to serve as an illustration of the "impure microkernel" idea discussed in the paragraph before it. The quote can be hedged further by making it more explicit that Torvalds is the maintainer of a competing product; I understand that not all readers will be familiar with his cocky way of arguing and might take the quote at face value. QVVERTYVS (hm?) 13:21, 13 October 2013 (UTC)

It doesn't belong in criticism either, since there is no technical information given. It's merely libel. TEDickey (talk) 13:23, 13 October 2013 (UTC)

Hm. Alright, I got rid of it. QVVERTYVS (hm?) 13:37, 13 October 2013 (UTC)

sounds good. If he had made a detailed technical criticism, that would be topical TEDickey (talk) 15:05, 13 October 2013 (UTC)

Citation for claim that all Windows operating systems from past a certain date are Windows NT?

I would like to see a citation supporting the following claim:

"Windows NT is a family of operating systems produced by Microsoft, the first version of which was released in July 1993 ... Windows 2000 and its successors are members of the Windows NT family, although they are not branded using the name 'Windows NT'." and for the inclusion of Windows 8.1.

I don't see any justification for calling every version of Windows until the end of time "Windows NT". I would have less of a problem if we differentiated between Windows NT the operating system and the Windows NT family of operating systems and if we used Category:Windows NT family of operating systems or even Category:Windows NT architecture instead of Category:Windows NT, but most of the references I can find for "Windows NT family of operating systems" are rather old and in the context of explaining how this new OS is different from the Windows 95 family of operating systems. I would very much like a citation to a reliable source that tells us which operating systems are or are not part of the Windows NT family of operating systems. --Guy Macon (talk) 22:16, 22 April 2014 (UTC)

Part of the problem here is that "Windows NT family" is not an official Microsoft term. We could probably find RS's for claims that Windows 2000 was a development from NT 4, XP was a development from 2000, etc., but I doubt you'll find a single blanket statement covering the whole "family." Jeh (talk) 22:41, 22 April 2014 (UTC)

btw, I support both Category:Windows NT family of operating systems and Category:Windows NT architecture, but I feel the latter should be reserved for articles that actually describe the internal architecture. Jeh (talk) 22:46, 22 April 2014 (UTC)

Good point about the internal architecture. I won't propose putting anything else in that category. --Guy Macon (talk) 23:29, 22 April 2014 (UTC)

Here you go, OP: http://msdn.microsoft.com/en-us/library/ms537503(v=vs.85).aspx Dogmaticeclectic (talk) 22:49, 22 April 2014 (UTC)

I really don't think that the fact that user agent strings (which, I might point out, call IE, Firefox, Chrome and even Googlebot "Mozilla") use the term "Windows NT" proves anything other than the fact that browsers tell lies to websites in order to avoid user-agent sniffing. --Guy Macon (talk) 23:29, 22 April 2014 (UTC)

Except that Microsoft has changed the user agent string for its browser significantly recently (as is indeed mentioned on that very page). If "NT" was no longer an actual designation, surely it would have been removed along with those changes? Dogmaticeclectic (talk) 23:31, 22 April 2014 (UTC)

Ignoring for a moment that the above is WP:SYNTHESIS and that Wikipedia needs citations to reliable sources, why would Microsoft do something that stupid? I use User Agent Switcher, and I have found that if my user-agent string doesn't contain "Windows NT" I often get a notice saying "you need to upgrade your browser to view this page". That microsoft page you cited explains in great detail how Microsoft IE, like every other browser, lies to web pages by putting false information in the user-agent string. For example, http://msdn.microsoft.com/en-us/library/bg182625%28v=vs.85%29.aspx clearly says that in IE11 "The navigator.appName property now returns "Netscape" to reflect the HTML5 standard and to match behavior of other browsers." and "The navigator.product property now returns "Gecko" in order to reflect the HTML5 standard and to match behavior of other browsers." Tell me, Is IE11 a rebranded Netscape Navigator? Does it use the Gecko (layout engine)? No? Then find a reliable source that establishes which parts of the user-agent string aren't bald-faced lies. --Guy Macon (talk) 00:03, 23 April 2014 (UTC)

The so-called Wikipedia:Synthesis in this case is actually not that at all, as it merely concerns a refutation of your argument, and not anything actually written in this article. In any case, though, Microsoft states the following at the link I provided, something that sounds clear enough to me: "Platform tokens describe your operating system." It seems to me that you should provide evidence (as opposed to similar synthesis-like arguments of your own) that Microsoft explicitly no longer considers NT to exist; failing that, the current wording of this article regarding this issue should remain. Dogmaticeclectic (talk) 00:45, 23 April 2014 (UTC)

The point that "Browser tokens describe your operating system" is not definitive. All a browser token really means is "if your site is compatible with a browser that advertises this token, your site should work with this browser." The quotes found by Guy Macon bear this out. This goes all the way back to MS's calling the first NT "NT 3.1" to match the then-current version of 16-bit Windows - so that 16-bit apps looking for that OS version would be happy. It most certainly was not the "three point one'th" version of Windows NT, there having been no version of Windows NT prior.

In any case, what Microsoft officially says about the matter may not be the most salient point. Maybe MS marketing doesn't want "Windows 8.1" officially connected to "Windows NT", but to anyone familiar with the OS internals, it is completely obvious that there is a clear line of development that runs all the way from NT 3.1 to Windows 8.1. We need to find RS's that state that. Jeh (talk) 02:05, 23 April 2014 (UTC)

"We could probably find RS's for claims that Windows 2000 was a development from NT 4". "Reliable" is ultimately going to be a matter of opinion (absent an algorithm, any characterization is going to be subject to interpretation), but there once was an operating system that Microsoft called Windows NT 5.0 but it had a different name when Microsoft released it.

There's definitely a codebase, API, and family connection running from NT 3.1 to Windows 8.1. I don't ultimately care what name Wikipedia chooses for the family, but I consider any claim that there's no family connection to be an extraordinary claim requiring extraordinary evidence. Guy Harris (talk) 03:41, 23 April 2014 (UTC)

I agree, to deny the family connection would be an extraordinary claim. It's one of those things that "everyone knows" but since MS has moved away from the NT branding it is not as easy as one would think to find sources. Jeh (talk) 04:41, 23 April 2014 (UTC)

I also agree. We do need to make it clear that there is a line of development -- that they are not the same, but are closely related. However, when we decide what to call that related group of operating systems, I don't think that we should choose a term that is [A] not familiar to most modern Windows users and [B] also the name of one particular operating system in that line of development.

How about this: Win16 Family, Win32 Family, and Win64 Family? --Guy Macon (talk) 06:51, 23 April 2014 (UTC)

Object. Win32 is the old name for the standard API on the NT family. It was still called Win32 on the 64-bit editions. I say "old" and "was" because recently MS is just calling it the "Windows API". The 64-bit versions of the NT family are compiled from the same source code tree as the 32-bit versions. Win64 isn't an official name of anything (though that term was bandied around a bit during the early development period). In any case, I don't think the 32-bit editions of the NT family and the 64-bit editions of the NT family should be grouped under different categories or descriptions. (Would we have yet another name for the ARM editions?) Jeh (talk) 07:13, 23 April 2014 (UTC)

And, for that matter, Win32 was also the old name for the mostly-Unicode-less (unless you had the Microsoft Layer for Unicode) version of that API on the "Windows OT" family starting with Windows 95. Guy Harris (talk) 07:54, 23 April 2014 (UTC)

Good points. Also, I just remembered that there was a Win32S available on Windows for Workgroups 3.11, so definitely scratch that idea. Looks like "Windows NT Family" is the best suggestion so far. Does anyone have a better suggestion? --Guy Macon (talk) 08:23, 23 April 2014 (UTC)

Hi.

This looks like a the "hyponymy of man" problem, where we have two types of "man" called "man" and "woman". Some solve it by calling the first man "human". Windows family divides into three actively developed subfamily: Windows NT, Windows Compact Embedded and Windows Phone (formerly Windows Mobile). There is the historical Windows 9x too. Although the Windows NT branch has become "the Windows" branch, "Windows NT" is not dropped: System.Environment.OSVersion object in Microsoft Windows still returns such strings as "Microsoft Windows NT 6.3.7601 Service Pack 1". (See [5] and [6])

So, yes, "Windows NT family" is plausible because we call each version the successor, the predecessor, the sibling or an edition of another.

Best regards,

Codename Lisa (talk) 13:35, 23 April 2014 (UTC)

Just for fun, there's also Windows RT where "Microsoft stated that Microsoft Store staff members would be given an average of 15 hours of training prior to the launch of Windows 8 and Windows RT to ensure consumers are able to make the correct choice for their needs." Okay, so we all need 15 hours of training to be able to tell these peas in the pod apart. :-) --Marc Kupper|talk 22:20, 11 September 2014 (UTC)

Guy Macon, Jeh, Dogmaticeclectic, Guy Harris, Codename Lisa, I decided to tackle the issue of finding a WP:RS that ties all of this into a "family". Microsoft's OSINFO structure's article says:

• VER_PLATFORM_WIN32s, or 0x0000, to specify Microsoft Windows 3.1.
• VER_PLATFORM_WIN32_WINDOWS, or 0x0001, to specify Windows 95, Windows 98, or operating systems descended from them.
• VER_PLATFORM_WIN32_NT, or 0x0010, to specify Windows NT or operating systems descended from it.

The use of the word "descended" is still vague on if we can call this a "family" but is getting closer. Microsoft's OS_TYPE Enumeration links the family a bit better in that it classifies machines as one of:

• VER_NT_WORKSTATION: The OS is a Windows NT workstation.<224>
• VER_NT_DOMAIN_CONTROLLER: The OS is a Windows NT domain controller.<225>
• VER_NT_SERVER: The OS is a Windows NT server.<226>

Click on <224> above to learn that a "Windows NT workstation" indicates "Windows 2000 Professional, Windows XP, Windows Vista, Windows 7, Windows 8, or Windows 8.1."

Click on <226> above to learn that a "Windows NT server" indicates "Windows 2000 Server, Windows Server 2003, Windows Server 2008, Windows Server 2008 R2, Windows Server 2012, or Windows Server 2012 R2."

I had also looked at the WIN32 API's GetVersionEx() function which returns an OSVERSIONINFO structure that contains the dwPlatformId field. This field is defined as holding the value 2 or VER_PLATFORM_WIN32_NT meaning "The operating system is Windows 7, Windows Server 2008, Windows Vista, Windows Server 2003, Windows XP, or Windows 2000." While that Microsoft article talks about Windows 8 and 8.1 it's silent on what the dwPlatformId is for those operating systems. I've submitted feedback to Microsoft to see if they will fix that detail.

I suspect Microsoft's OS_TYPE article and its footnotes are decent WP:RS that Windows NT and its descendants up through Windows 8.1 / Windows Server 2012 R2 are in the same family. I'm not sure how to work that into the Wikipedia article without a bunch of hand waving. I'd like to "show" rather than "tell" the WP reader it's a family.

In my own mind I've always called it the "Windows NT architecture" rather than a "family" but in reading over this thread it seems "family" is a better word and that we use "Windows NT" to separate this from the older generations of the Microsoft Windows family. --Marc Kupper|talk 22:20, 11 September 2014 (UTC)

• Comment - I have removed the hat notice about this discussion thread from the article. While active discussion ended in April it's not clear if there was a consensus other than perhaps that the article title remains at "Windows NT" and calling it the "Windows NT family" in the article is the lesser of evils. --Marc Kupper|talk 19:48, 11 September 2014 (UTC)

Requested move 26 April 2014

The following discussion is an archived discussion of a requested move. Please do not modify it. Subsequent comments should be made in a new section on the talk page. Editors desiring to contest the closing decision should consider a move review. No further edits should be made to this section.

The result of the move request was: not moved. -- BrownHairedGirl (talk) • (contribs) 17:54, 8 May 2014 (UTC)

---

Windows NT → Windows NT family (typo corrected; was "Windows NT Family") – This rename will differentiate between Windows NT the specific operating systems named "Windows NT" and Windows NT the family of related operating systems that includes Windows XP, Windows 7, and Windows 8. -- Guy Macon (talk) 20:55, 26 April 2014 (UTC) Edited 12:04, 27 April 2014 (UTC)

Survey

Feel free to state your position on the renaming proposal by beginning a new line in this section with *'''Support''' or *'''Oppose''', then sign your comment with ~~~~. Since polling is not a substitute for discussion, please explain your reasons, taking into account Wikipedia's policy on article titles.

• Oppose thrice over. It's not a family; the people who live next door are a family. Regardless, we would not capitalize it like that. Third, there is no "Windows NT the specific operating system"; there are Windows NT 3.1, etc. each with their own articles, so there is no naming or disambiguation problem here at all. This article is a very well-develoepd set index article and serves its function just fine, at its present name. Compare Dodge Charger.  — SMcCandlish ☺ ☏ ¢ ≽^ʌⱷ҅_ᴥⱷ^ʌ≼  23:34, 26 April 2014 (UTC)

• Just to be clear, you don't have any problem at all with calling Windows 8 "Window NT" instead of calling it "part of the Windows NT Family of operating systems"? Guy Macon (talk) 05:00, 27 April 2014‎ (UTC)

I, at least, don't have a problem calling it Windows NT 6.2. Guy Harris (talk) 06:17, 27 April 2014 (UTC)

Thanks for making that clear. I have no problem if the consensus goes against me, but the "Dodge Charger" comment above made me suspect that some here might not understand that they are not just doing the equivalent of calling both the 1964 Dodge Charger concept and the 2014 Dodge Charger LX a "Dodge Charger" but also calling the Dodge Neon and Dodge Sprinter "Dodge Chargers". Of course no analogy is perfect, and Chrysler doesn't call a Sprinter "Dodge Charger 6.2" in their service manual... Like I said, if the consensus is to call Windows 8 "Window NT" then I am fine with following the consensus, but I would prefer it if I could find a reliable third party source that calls it that.

As for the claim "It's not a family; the people who live next door are a family", the article starts with the words "Windows NT is a family of operating systems...". I don't think anyone has an issue with the article saying that; the only issue here is what to title the article --Guy Macon (talk) 12:04, 27 April 2014 (UTC)

If the Neon and Sprinter are just rebranded variants of the Charger it would certainly be appropriate to mention them as such at the Dodge Charger article.  — SMcCandlish ☺ ☏ ¢ ≽^ʌⱷ҅_ᴥⱷ^ʌ≼  09:08, 6 May 2014 (UTC)

Windows 8 is just as much a rebranded variant of Windows NT as the Neon and Sprinter are rebranded variants of the Charger. Clearly they are all of the same family (reciprocating internal combustion engine, water cooled, four rubber tires, glass windscreen with dual wipers, body made of painted metal...) but just as clearly not the same thing and should not all be all be called "Chargers". While the Neon, Sprinter and Charger are of the same family, a Harley Davidson or a Jet-Ski are of different families. Likewise, Windows 8 has far too many differences from Windows NT to be called a "rebranded variant". --Guy Macon (talk) 12:01, 6 May 2014 (UTC)

• Oppose. As per SMcCandlish's comments, Windows NT is already the name of the family. pcuser42 (talk) 04:40, 27 April 2014 (UTC)

• Oppose (2x). First, the MOS:CAPS problem: "Family" needs to be "family"; (Fixed) Second, we must not lengthen the name unnecessarily. The proposed rename is analogous to proposing Bill Gates article be renamed "Bill Gates (Human)". Third, there is no "Windows NT the specific operating system"! Windows NT does not even have a version 1.0! Best regards, Codename Lisa (talk) 05:33, 27 April 2014 (UTC)

• Fixed capitalization typo. Sorry about that. --Guy Macon (talk) 12:04, 27 April 2014 (UTC)

• Thanks Wikipedian Guy Macon, human. Still, no. Best regards, Codename Lisa (talk) 12:28, 27 April 2014 (UTC)

• Oppose: Windows NT refers to the family of operating systems and nothing else. Adding a word to the title would be unnecessary. --TheMillionRabbit 17:16, 27 April 2014 (UTC)

• Weak support: The purpose of an article title is to allow people to find the article if they are looking for the information the article contains. This article is about the entire NT family, from the pre-release development of NT 3.1 through today's 8.1-with-update and Server 2012 R2, and that is fine: I think WP does need an article with this content. But I don't think "Windows NT" is a title that will help the general reader find that content.

Will someone who is looking for info on "the family of operating systems that includes Windows 7" know to type "Windows NT" into the search box? Whatever they search for, if they see "Windows NT" in the results, will the general reader think such an article is applicable to Windows 7? I doubt it. They'll think "oh, I'm not running NT, is anyone these days?" and look at other search results. Will adding "family" to the name help? I think it will help guide some readers here, hence the weak support.

However: I think that most general readers, if they are aware of NT at all at this point, are aware only that it's an old name for a "Windows-like OS" that they have never actually touched. A smaller group, but still larger than those who are aware of the true nature and extent of what we are calling the "NT family", likely think that the family's last member was NT 4.0.

That Microsoft's last official use of the "NT" moniker was in Windows 2000's tag line does not help.

Some of you are no doubt about to reply "but 'NT family' is technically correct!" Believe me, I am aware of this! The line of descent from NT 3.1 through Windows 8.1-with-update is completely apparent to "insiders", and there is no reason to think that anyone taking an interest here is not in that category. So, sure, we know what the article "ought" to be named, for good technical and historical reasons.

But that makes it tough for us to see things as the general reader sees them. I make a good part of my living teaching and consulting on Windows internals and related subjects, and I can tell you that even among IT professionals, the fact that there is a single chain of development from NT 3.1 to Windows 8.1 is a surprise to at least a fourth. (I've heard comments of the form "I thought they rewrote everything from scratch for Windows x", for almost all possible values of x; x=="Vista" is the most common. And, no, they didn't. :) )

As for the "general reader" for whom WP is written? I would bet that at least half of them think that Windows XP, Windows Vista, perhaps Windows 7, and certainly Windows 8 are each "whole new operating systems."

Summary: I don't think the general readership is more than dimly aware of "Windows NT", let alone the "family" nature of NT 3.1 through Windows 8.1+. So either article title, "Windows NT" or "Windows NT family", seems to me to violate WP:COMMONNAME. COMMONNAME also notes that article titles should not be pedantic, and I think these titles at least border on that. Having said that, I have to add that I can't think of anything better! I do think that "Windows NT family" conveys the content to the general reader better than just "Windows NT".

But I think the larger problem of "how to help the general reader find this information, or recognize that this article is of interest to them" is not well served by either name. Jeh (talk) 20:16, 27 April 2014 (UTC)

Perhaps the general reader (one unaware, or only minimally aware, of the transition from "Windows OT" - 1.x, 2.x, 3.x, 95, 98, and Me - to "Windows NT") is best served by having the Microsoft Windows article lead them here if they're interested, and perhaps more strongly indicate the membership of W2K and post-W2K releases in the family. Perhaps the "Version history" section should split into three sections for the three code bases (OT, NT, and CE), or perhaps there should be a separate section discussing that. Guy Harris (talk) 21:07, 27 April 2014 (UTC)

---

The above discussion is preserved as an archive of a requested move. Please do not modify it. Subsequent comments should be made in a new section on this talk page or in a move review. No further edits should be made to this section.

Windows Phone

Recently all versions of Windows (XP+) have been changed to Windows N.T. ¿Should Windows Phone 8 and later also have this on the top of their infoboxes? as they are also based on the Windows N.T. kernel ¿or are they different because of other reasons? Sincerely, --Namlong618 (talk) 16:45, 18 February 2015 (UTC)

(Presumably you mean "the infoboxes for all the NT-based versions of Windows have recently been changed to say "A version of the Windows NT operating system" rather than "A version of the Microsoft Windows operating system"; it's Wikipedia, not Windows, that changed, and the change goes back to the Windows NT 3.1 page.)

That question is probably best asked on the Windows Phone talk page, as it's a question about those versions of Windows Phone (i.e. "why aren't these put into the same category as the desktop/laptop version of Windows?"), not a question about Windows NT as a whole. Guy Harris (talk) 19:15, 18 February 2015 (UTC)

Actually, no, they shouldn't. Members of Windows NT family all report "Windows NT" with their version number API. Windows Phone members do not; they don't even have the API.

Best regards,

Codename Lisa (talk) 18:01, 19 February 2015 (UTC)

Putting Windows Server information in an infobox?

The Windows NT family includes mobile, embedded, desktop/notebook, and server OSes. The infobox's release information is for the desktop/notebook flavor; should there be a separate infobox for the server flavor, and perhaps also on the Microsoft Windows page, or should that go on the Windows Server page? This page doesn't give a marketing target in the infobox; the Microsoft Windows page does, giving it as "personal computing", but it also covers non-desktop/notebook flavors of Windows. Guy Harris (talk) 23:20, 19 July 2015 (UTC)

Vista disk space increase primarily due to DWM memory usage?

Current version says that the 10-fold disk space increase in Vista is primarily due to the extra memory needed by the Desktop Window Manager. This doesn't sound right -- it's mixing system memory / VRAM with disk space. The Vista DWM was notorious for using a lot of memory and causing slow performance as a result -- which was fixed by rearchitecting it in Windows 7 -- but I've never heard of it being a significant cause of disk space usage and no citation has been provided for this claim. 24.130.133.67 (talk) 09:18, 22 July 2015 (UTC)

Removed. Additional memory usage would only increase disk space usage to the extent that the memory needs backing store on disk, and unless WXP required only a tiny amount of disk space, whatever increase that causes in pagefile size isn't going to be a 10x increase. Guy Harris (talk) 17:13, 22 July 2015 (UTC)

Windows 10 versioning

Windows 10 is actually 6.4 despite the 2014 news stating that it will be 10.0. Someone should clean up the article. ebertek (talk) 15:53, 24 March 2016 (UTC)

Early preview versions used the number 6.4 but later previews and the final version use 10.0. - Josh (talk | contribs) 18:32, 24 March 2016 (UTC)

Northern Telecom

Shouldn't there be a mention that NT was bought from Northern Telecom? — Preceding unsigned comment added by 67.78.5.254 (talk) 15:45, 8 April 2016 (UTC)

Not without a citation. (Presumably you mean that Microsoft bought the rights to use "NT" in their OS name from Northern Telecom, similar to Apple having to get the rights to use "iOS" as an OS name from Cisco.) Guy Harris (talk) 16:58, 8 April 2016 (UTC)

Hi. Stress on the source part because the assertion is dubious. I have already looked in the U.S. trademark database (tmsearch.uspto.gov)and it seems Microsoft has owned the "Windows NT" mark; stress on "Windows" part.

Best regards,

Codename Lisa (talk) 14:55, 9 April 2016 (UTC)

Multiple version numbers in the infobox?

There is currently a discussion of whether Template:Infobox OS should be used with multiple version numbers - for example, to list both a "software update" and "next major release" beta, or to list betas from more than one release stream. If you believe that multiple {stable, preview} releases should never appear in that infobox, or if you believe that they should appear under some or all circumstances where there's more than one beta of the OS in question available, you might want to comment there. (I have no strong belief either way; I'm OK with the main OS page listing only the "next major release" beta, but listing betas from multiple streams if they exist, but I'd also be OK with other choices.) Guy Harris (talk) 08:09, 11 July 2016 (UTC)

NT 4 kernel source code on GitHub

Please see Talk:Windows NT 4.0 § NT 4 kernel source code on GitHub —Codename Lisa (talk) 07:07, 21 May 2017 (UTC)

Page protection?

Not that I want move this page to a new title, but is there any reason it is still move protected over a single instance of page move vandalism from 8 years ago? It doesn't seem to meet the purposes set out in WP:MOVP. Kb.au (talk) 14:42, 30 June 2017 (UTC)

Hello.

Did you read the third reason on the page to which you linked?

Highly visible pages that have no reason to be moved

This article is such a page. In applying all Wikipedia policies, common sense is a requirement. In this case, the decision to protect it is just good sense.

Best regards,
Codename Lisa (talk) 15:16, 30 June 2017 (UTC)

Non-encyclopedic language in overview section

I think the following paragraph (last in the overview section) uses encyclopedic language, and could be unnecessarily promotional, but I didn't want to make changes since I don't often visit this page and am not familiar with the history:

"In some ways the ancestors of Windows NT are the operating systems that Dave Cutler worked on before he was hired by Microsoft, namely the VMS and RSX-11 operating systems, and also an unreleased object-based operating system developed by him for DEC Prism. This lineage is made clear in Cutler's foreword to "Inside Windows NT" by Helen Custer.[4]"

The phrase "in some ways" is clearly nonspecific and those "ways" should probably be enumerated. Also, I think the general tone of this sounds a little promotional towards Dave Cutler (he is the author of the source cited to prove the claim that he developed the "ancestors" of Windows NT. If he is really a big figure in Windows NT that's obviously fine, but this sentence doesn't make clear what his contributions were, or why they were important enough that he is listed in the overview section rather than the history section.

Basically, when reading this section I felt kind of like I was talking to some weathered Computer Scientist giving me his version of how Windows came to be, not like I was reading an objective encyclopedia. — Preceding unsigned comment added by Ludicous (talk • contribs) 16:45, 8 September 2017 (UTC)

Your time would probably be better spent fixing it there than complaining about it here... Warren -talk- 00:06, 9 September 2017 (UTC)

Taken care of! —Codename Lisa (talk) 03:39, 9 September 2017 (UTC)

Wrong Windows Logo

Why the new Windows logo is being used for this 1993 released OS? BarricadeX (talk) 04:00, 14 August 2018 (UTC)

"Windows NT" here refers to a family of OSes built on the same software architecture, not to a particular member of that family; the first OS in that family happened to have the string "Windows NT" in its name, but the current OSes in that family have strings "Windows 10" and "Windows Server 2016" in their names. A number of different logos have been used for various members of that family; we're currently using one of the more recent logos. Guy Harris (talk) 05:20, 14 August 2018 (UTC)

For the Windows NT operating system released in 1993, see Windows NT 3.1. pcuser42 (talk) 08:30, 14 August 2018 (UTC)

Alright, thanks for the info. BarricadeX (talk) 16:57, 14 August 2018 (UTC)

Contested deletion

This page should not be speedy deleted as pure vandalism or a blatant hoax, because I WANNA DELETE THIS PAGE — Preceding unsigned comment added by Ziadelwan (talk • contribs) 19:08, 29 August 2018 (UTC)

A Commons file used on this page or its Wikidata item has been nominated for deletion

The following Wikimedia Commons file used on this page or its Wikidata item has been nominated for deletion:

• Windows NT Workstation 4.0 Installation Disc.jpg

Participate in the deletion discussion at the nomination page. —Community Tech bot (talk) 10:40, 30 June 2021 (UTC)

Official statement on NT being a hybrid kernel?

2401:D800:2220:8C36:3152:2512:9718:BA30 (talk · contribs · WHOIS) / 2401:d800:2de0:e28e:56fd:f712:2d14:f7fa (talk · contribs · WHOIS) (I'm assuming they're the same person) has recently tagged this page – and hybrid kernel – with {{Citation needed}}, stating that we need official confirmation from Microsoft to call NT a hybrid kernel. I don't see how that is grounded in policy, though. WP:V requires reliable sources, so if reliable sources call NT a hybrid kernel, that should be stated. I don't know of any policy that would additionally require confirmation by the vendor of a product in a case like this. ;; Maddy ♥︎(they/she)♥︎ :: talk  08:30, 24 August 2022 (UTC)

Regardless of whether confirmation of the OS vendor is necessary, there's no peer-reviewed credible sources that say NT is a hybrid kernel. There's also no credible source that even define what a hybrid kernel is in the first place (so the whole definition section in Hybrid kernel would now need citation from actual credible sources, not osdev.org and not techopedia, which is definitely not a reliable source), so the tags will stay for the meantime. 2401:D800:2DE0:E28E:56FD:F712:2D14:F7FA (talk) 09:10, 24 August 2022 (UTC)

I'm in Camp Torvalds on this issue; without a sufficiently reliable source indicating what a "hybrid kernel" is, any designation of an OS as having a "hybrid kernel" may just be One Person's Opinion.

Hybrid kernel § Overview says:

The idea behind a hybrid kernel is to have a kernel structure similar to that of a microkernel, but to implement that structure in the manner of a monolithic kernel. In contrast to a microkernel, all (or nearly all) operating system services in a hybrid kernel are still in kernel space. There are none of the reliability benefits of having services in user space, as with a microkernel. However, just as with an ordinary monolithic kernel, there is none of the performance overhead for message passing and context switching between kernel and user mode that normally comes with a microkernel.

which spends most of its time saying how a "hybrid kernel" differs from a microkernel, with the only similarity being that it has "a kernel structure similar to that of a microkernel" without indicating what that "kernel structure" is. Guy Harris (talk) 09:24, 24 August 2022 (UTC)

I notice hybrid kernel actually does cite two Microsoft sources. At least the first one (I can't access the second since I don't have PowerPoint installed on this machine) refers to it as a "modified microkernel" or "macrokernel" but I think from the way they describe it, we can reasonably include NT under that article – with explanation of course.

Regardless, I would still change your tags if we leave them. On Windows NT I'd remove the text about it needing to be from MS, and on hybrid kernel I'd do the same plus either inline tag specific statements you think are problematic, or use another template which is meant to tag an entire section. ;; Maddy ♥︎(they/she)♥︎ :: talk  09:25, 24 August 2022 (UTC)

The first reference says, in the first paragraph of the "Microkernel Designs and the Windows NT Architecture" section:

The Windows NT 4.0 architecture merges the best attributes of a layered operating system with those of a client/server or microkernel operating system. The Windows NT Executive is the generic name for a number of subsystems and components of the operating system that execute in the microprocessor's kernel (or privileged*)* mode. The remainder of the operating system, such as various server subsystems and their application clients, execute in user (or application) mode.

}

It later makes a bit of a surprising statement:

A microkernel, on the other hand, is the name given to the core portion of a modern, modular operating system. Microkernel operating systems are based on two fundamental principles. The most basic principle is one of modularity, encapsulation, and data hiding. In this aspect of the design, there is one and only one portion of the operating system that has system-wide responsibility for a particular function. All other parts of the operating system (as well as applications, naturally) access that function through a well-defined interface. There is no duplication of function and no "back doors" to critical data structures; all access to system data structures is through software interfaces. This approach makes it possible to upgrade or replace entire modules within the system without disturbing the remainder. A secondary principal of the microkernel design, related to the first but focused more on the implementation strategy, is that large portions of the operating system which traditionally run entirely in the kernel or privileged mode of the microprocessor can now be executed in user or application mode, with only the microkernel itself, along with a relatively small amount of hardware device-specific code, executing in kernel mode.

saying that a "microkernel" having a modular design is more fundamental than a "microkernel" moving a large amount of functionality to user-mode code.

That means that the main principle of a "microkernel" also applies to cleanly-designed privileged-mode code in an OS where a lot of functionality is implemented in privileged code. I suspect that, at this point, a lot of conventional UN*X kernels also follow that principle to a large degree; they don't indicate whether the "back doors" to critical data structures are enforced by anything more than an agreement by developers not to touch other modules' data structures except by calls to those modules' code.

(The one place that paper says where NT "has, until now, followed the more pure microkernel approach" is "its implementation of the window manager and graphics subsystems which implement the GUI aspects of the Win32 API", in which the window system is implemented in the user-mode "Client-Server SubSystem" (CSRSS) process, with graphics drivers running in that process. All I have to say to that as an example of a "pure microkernel approach" is "the X Window System says hi!".)

So the NT paper seems to think that "microkernel" primarily means a kernel that provides services to other kernel-mode components, providing a framework into which kernel-mode drivers, networking stacks, file systems, etc. can be plugged, not a kernel atop which that functionality is implemented by user-mode processes, with the graphics subsystem being the only place in NT (prior to 4.0) where a user-mode process is a critical component. I'm not sure to what extent that differs from a boring old monolithic kernel with loadable kernel modules, as the loadable modules, not being compiled and linked together with the monolithic kernel, have routines in the kernel that they're supposed to call to get various services, including registering routines in the module to be called by the kernel code.

The one place where the NT appears to differ from an OS with a boring monolithic kernel with loadable modules, using X11 as its window system, appears to be that some APIs that would be thought of as "system calls" involve the aforementioned user-mode CSRSS process (or another subsystem's user-mode process). That continued to be true even after the graphics subsystem was moved to kernel-mode code; the PowerPoint slide reference mentions the "NT Subsystem model" and notes that process creation involves CSRSS, the Architecture of Windows NT mentions it, and Client/Server Runtime Subsystem is the Wikipedia page about it but doesn't give a lot of detail. (I have the impression that CSRSS might also manage the drive letters in pathnames, with CreateFile() sending a message to CSRSS, which then constructs a path name that it then passes to the NtCreateFile() system call, the result of which it sends back.)

As for Darwin, until recently I'm not sure what distinguished it from an OS with a boring monolithic kernel with loadable modules; it doesn't even have an equivalent to CSRSS. For example, open() is a system call, unlike CreateFile(), and fork(), the exec calls, and posix_spawn() are system calls, unlike CreateProcess(). I'm not sure how its graphics/window system layer currently looks, but I suspect it may not be too different from current Linux systems using Wayland. With DriverKit, some drivers - including, from my experience with Ethernet driver crashes when running macOS Ventura on a VMware Fusion virtual machine, at least one Ethernet driver - may have been moved from the kernel to user mode processes.

Some editor insists on labeling Oracle Solaris as a hybrid kernel, without providing any evidence for how it's different from an OS with a boring monolithic kernel with loadable modules. The reference given for OS/2 being a hybrid kernel,, and the "Inside the OS/2 Kernel" page to which it links, don't seem to give any such evidence, either, e.g. the file systems appear to be Installable File System kernel loadable modules.

And the Filesystem in Userspace and Rump kernel mechanisms make the whole question even more complicated.

(BTW, the term "micro" in "microkernel" appears to be doing a lot of work here.

The paper cited by microkernel as an early use of the term use the term "MicroKernel" to refer to part of the [[microcode] on the PERQ computers on which the Accent kernel was being developed, with "micro" here referring to it being in microcode; at least as I read the paper, there's also a non-microcode kernel that provides the sort of services one might expect of a "microkernel" in the now-common sense.

The Microsoft paper appears to view "micro" as primarily meaning "not including all kernel-mode code, but providing core services to other kernel-mode code", with "not having much non-kernel-mode code in the first place, as that functionality has been moved to user-mode servers" as secondary.

Most other uses seem to mean "provides services to user-mode servers, which implement functionality such as device drivers, network stacks, and file systems".) Guy Harris (talk) 20:17, 24 August 2022 (UTC)

References