First ever 64-bit version of Windows rediscovered and a C compiler for it too | Hacker News Reader

First ever 64-bit version of Windows rediscovered and a C compiler for it too

theregister.com

96 pointslproven28 comments

> in fact, based on the Windows Server 2003 codebase rather than XP per se

For a while in the 1990s, Microsoft tried to pretend that "Server" and "Workstation" NT were substantially different. At some point somebody actually worked out which registry key actually matters, and then (though obviously it's unsupported) you could turn a "Workstation" into a "Server" or vice versa, because in fact, just like Linux it turns out an operating system kernel is an operating system kernel, marketing is bullshit, the difference is always just tuning.

After that, Microsoft didn't ship "Server" and "Workstation" variants of the exact same code. Instead they'd branch the "Server" and "Workstation" products, under different names, from the same tree at different intervals. Thus this difference is purely about the badge, and presumably they felt that the XP badge was appropriate for this product whereas a "Server" badge would give the wrong impression.

Nobody had to "work this out" because it was widely known and acknowledged by MS since the first AS beta release (1993). The original idea was to sell a product to complete with Novell Netware (at the time the dominant thing you ran on the office file server). If MS made it easy to use the desktop NT as a file server, then they'd be "leaving money on the table" vs what Novell charged for Netware (per-client licensing that meant $thousands for a large server). So they invented a license for clients (CAL) and a product that bundled some number of licenses in the SKU, and AS was born, sold for iirc about $500. Much cheaper than Netware, which subsequently died. Then, to make it harder for customers to simply use the cheaper desktop NT in place of AS, a limit on the number of connected clients was enforced (5?). There were also some kernel tuning parameters that were different between workstation and server.

Subsequently, through the 2000, 2003, ... generations, more applications were bundled into the server product (e.g. ActiveDirectory) that were obviously not present in the desktop version.

When you say "acknowledged by Microsoft" I notice you don't provide a link or even a quote. There's a big difference between what's obvious to programmers and what you can get fanboys to accept is true about a product they've somehow connected to their self-worth. In my experience the Registry Key made all the difference and of course Microsoft were not publishing "Use this registry key to effectively pirate Windows Server" docs - unless you're claiming they did?

Windows 2000 Server is like older NT versions, it's exactly the same actual code as the far cheaper Windows 2000 Professional, whereas Windows Server 2003 was a different branch from the similar XP on the desktop, there was no "Windows Professional 2003" or whatever. The bundling makes it a bit easier for customers to stomach that you're blatantly selling them the exact same product with a different badge for more money, but it's not enough.

It wasn't just about Client licenses either. Some Windows "editions" (ie the exact same code with a registry change) had different hardware support. Got 4 CPUs? Can't use them in "professional" edition, you need Server. Got a huge beat with 32 CPUs? Ordinary Server won't work, get yourself Datacenter Server. Same for RAM.

Windows XP and Server 2003 would mark the only time in Windows's history that the releases weren't fully synced up. (Unless you count Windows XP x64 Edition, which was the same OS files as Server 2003)

I doubt any Windows administrators are really mystified by it. Vista and Server 2008, one and the same. 7 and Server 2008 R2, one and the same. 8.0 and Server 2012, one and the same. etc. It's even obvious if you pay attention to update files, they start with "Windows6.0-KB...", "Windows6.1-KB...", etc.

As for marketing, there is a real difference between home/workstation releases and server releases. It's not in the kernel or basic OS services, but instead with things like Windows Media Player or IIS. It makes little sense to include WMP on a server. Likewise, it makes little sense to include IIS on Home Edition.

I've never understood how the difference between the editions being just a matter of having different bits set means "marketing is bullshit". How's changing the bits post-install different from, say, having Workstation vs. Server come on different CDs with those bits flipped? It's all bits in the end.

There's more difference than timing of when they cut a release. Different SKUs are crippled in different ways with features actually removed. And you can pay extra for upgrades to less crippled crippleware.

IANAL, so I'm surprised the RICCO Act doesn't apply.

Mark Russinovich wrote about this back in 1996: https://www.itprotoday.com/compute-engines/inside-difference...

I don't know if modern versions of Windows will still blue-screen when you attempted to make changes to the "ProductType" and "ProductSuite" REG_SZ values in the "HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\ProductOptions" key. (I see some values under that key in Windows 10 that I don't remember seeing in older Windows versions, too.)

I don't think it was ever a secret that the kernel and system was the same lineage, it was all about the config and whether various optional things (like IIS) were there by default.

NT and 2000 were always business-targeted products with all the attendant usage-based licensing stuff that wasn't present in the consumer world, even though 2000 had a wide enthusiast userbase that was tired of 98 and ME and mostly figured out how to get games running well enough.

Apple and various linux distros separated their products into "server" and "workstation" variants in the same way, too.

> At some point somebody actually worked out which registry key actually matters

That “someone” was Mark Russinovich. He is now the CTO of Azure.

https://en.wikipedia.org/wiki/Mark_Russinovich

Discussed a couple of days ago at the original blog link (92 points, 76 comments): https://news.ycombinator.com/item?id=35949245.

Fair point. I submitted this because I wrote it, and in the story, I tried to add context and background that wasn't in the original.

Most people read the headline, don't even understand that, and then comment. :-(

> Most people read the headline, don't even understand that, and then comment. :-(

Welcome to HN.

I mean welcome to the internet, you see this happen even on news articles themselves, people will read the headline, scroll to the bottom to comment something that’s explained in the article.

For a long time the verge has users who were notorious for this. It got so bad that they had to institute a “summer of chill” where most articles had comments turned off (which was a novel concept at the time).

I'm aware. My primary email is 32 years old now, but I was online 6 years before that.

It took your article for me to go trawling in the blog’s comments section and realize (some of) the files were actually released[1], not just blogged about, so thanks for that :)

[1] https://archive.org/download/splash_20230516

Oh, good find! I did not spot that either.

Windows may not have taken advantage of the 64-bit CPU in these machines, but other operating systems of the time certainly did.

OpenBSD/alpha was a 64-bit only port (LP64) to the DEC Alpha machines, started in 1996 and remains supported today. OpenBSD/alpha even made several of the early printed CD-ROM releases and can be seen on the cover art.

https://www.openbsd.org/alpha.html

https://www.openbsd.org/artwork.html

https://cdn.openbsd.org/pub/OpenBSD/7.3/alpha/INSTALL.alpha

Alpha was also the first 64-bit processor Linux supported. AltaVista was originally just a tech demo for Alpha. The homepage, at least early versions, had little "AlphaPowered" graphics on it.

[Article author here]

This is true but understates the truth.

Alpha was the first non-x86 CPU of any kind that Linux supported.

I've used XP Pro x64 for several years on Pentium D CPU with full 4GiB of memory. It was rather pleasant experience, with one caveat: some desktop software refuses to sun because they detected server OS, and vice versa. But there were not so much so picky software, so it was not a real problem.

The "server OS" bit came from software assuming that Windows NT 5.2 would always be Server 2003, and it was an incorrect assumption. If they had instead read the ProductType key and based the SKU lockout on WinNT (home/workstation) or ServerNT (server, obviously), no such issue would have happened.

Windows 2000 was always NT 5.0 in both professional and server, such software had to use ProductType instead. Likewise, Windows Vista and Server 2008 are both NT 6.0. XP/Server 2003 is the only time when it deviated, and some 3rd-parties didn't notice the proper way to distinguish between client and server; becoming even worse once XP x64 came out and was actually NT 5.2 (XP 32-bit was NT 5.1).

Pretty cool. I wish copies of \\juliet were available, and whatever was used in the porting lab. \\juliet was a machine rescued from pc-recycle that a co-worker in ITG turned into an "everything" machine for network installs of NT

Not the oldest--we had Win NT 3.x (3.1 I think, maybe 3.5) for the DEC Alpha where I worked.

The older Alpha versions of Windows were mostly 32 bit even though the Alpha was natively a 64 bit CPU (see https://devblogs.microsoft.com/oldnewthing/20170823-00/?p=96...). This is a prototype of actual 64 bit Windows, early in development when actual Itanium silicon didn't exist, so they used Alpha as a test target even though Alpha Windows was dead.

That actually seems fairly interesting and unusual—in Linux terms, Windows NT (and its progeny) started with the moral equivalent of the x32 ABI[1] on its only 64-bit architecture (Alpha) and only afterwards introduced support for true 64-bit apps. Which is very reasonable!

[1] https://sites.google.com/site/x32abi/

Did you RTFA? I wrote it.

All released NT versions are 32-bit only.

> Please don't comment on whether someone read an article. "Did you even read the article? It mentions that" can be shortened to "The article mentions that".

https://news.ycombinator.com/newsguidelines.html

Running on 64 bit CPU != 64 bit OS

Reply on news.ycombinator.com