How Novell Netware lost the battle against Windows NT (2013)
theregister.com
theregister.com
It’s a startling deficiency that killed its chances of becoming a general purpose server.
Without these things, the Netware server couldn’t be trusted to run independent applications.
There were some third party server applications built for Netware, I think maybe CC:mail server and Lotus Notes, but the memory protection and cooperative multitasking meant they didn’t last long.
I don’t know if this is true but I heard it was Drew Major who made this call, which IMO killed Netware.
The justification I heard at the time was that memory protection was too slow. Better slow than dead. I think Novell just didn’t have the technical chops to do it.
Also Microsoft was at the height of its powers as a competition killing machine and nothing that stood in its way survived. It really genuinely looked back then like Microsoft was going to wipe out everything and it would be a Microsoft only future. Novell was no exception along with Lotus and Borland and many others.
But on the original topic, yes once you've got a 286+ there's no excuse for your OS not having true memory protection.
Despite that Mac continued with only partial process protection up until the release of OSX in the early 2000s.
My point is their problem wasn't hardware support, but software compatibility.
The newer Mac II's (IIx onward) were 68030-based, so they would've had MMUs built in. (They also released the Mac LC later on, which was 68020 based and had no socket for an MMU.)
Was Win95 preemptive or cooperative? I totally forget. Pretty sure only the NT versions of windows were preemptive.
I recall there was a period of time when you’d run Windows 2000 because it was so much more stable than 98/ME. It worked mostly okay but it didn’t have as good of support for gaming and was pretty late to the party with DirectX support.
They obviously had to make compromises for performance and compatibility’s sake that made the long term move to NT essential, but I feel like the Chicago kernel team’s efforts have never (to my knowledge) been well-documented or -appreciated.
A/UX ran with memory protection on the 68020 Mac II in 1988, probably in 1987 in beta also.
The book is "In Search of stupidity" by Merrill R. Chapman, this particular topic is covered on page 173-174.
except for the thousands of companies, academics and individuals who said NO
There was mostly a lot of appetite from both buyers and application vendors to standardize at the operating system level--which Unix in the mid-90s decidedly was not. There was an effort to come up with a multi-vendor Unix, which SCO predictably screwed up in their pre-lawsuit guise. Linux would eventually mature sufficiently and gain business support for this role on servers but that didn't really happen in a major way until the 2000s, accelerated by IBM's endorsement.
I worked in the heart of the corporate world and at this time the sentiment has changed from "you can't get fired buying IBM" to "you can't get fired buying Microsoft".
As @gaff said in this thread said - it has nothing to do with Microsoft ad spend or what the magazines said - it was dead obvious. And were it not for Linux then it would have been a Microsoft world. Remember also it was Microsoft that saved Apple from bankruptcy when jobs returned. I do think that if not Linux then FreePSD would likely have become todays "Linux".
[0] https://bitmason.blogspot.com/2020/06/podcast-was-open-sourc...
MSDOS on the PC was clearly not suitable for real business apps. Well not for networked ones. Well not for ones that need a server. Well not for GUI apps. Well not for apps that needed more powerful CPUs than those you x86 chips.
Wait.. why are all our customers on Windows NT on Pentiums?
One of the major SQL database vendors (maybe Sybase?) actually ported their product to NetWare. But no one bought it due to the problems you stated.
The NetWare development tools were also terrible. Whereas the early Windows NT development tools were actually pretty good. Unlike with NetWare, NT allowed developers to code on a workstation platform and then deploy on a server platform with an identical API which was a huge factor in getting third-party developers onboard.
On the other hand why hasn't anyone implemented a recycle bin like they did? (Your whole drive was your recycle bin.) And the users really loved the feature that said "this drawing is locked by <station name>" that I can see no way to implement under any other system.
There was a guy at our office that was extremely proficient with Netware, so much so that if he got physical access to a Netware server, he could press the special 'Admin key sequence' (from memory it was about 4 keys pressed at once on the keyboard) and then he would issue commands to peek into memory and tell you the Admin password. This trick only worked if you had already unlocked the screen console with the Admin account since the server booted, but it showed us very early on how important physical security to even a keyboard was.
I do remember a key fight that happened back then with Netware and Microsoft was how Netware gave you a free 2 user license to a server if that server was only running network services, such as maybe a DHCP server. If you wanted to do file sharing, then you had to buy the appropriate user license.
Microsoft introduced the ability for a Windows NT server to mount a share locally where the contents of that share was data from a Netware server. This way you could get a free 2 user Netware server license, have the NT server use one of those licenses, and then hundreds of people could access the data without Netware being able to technically block it.
I recall (can't find any images on the Internet to back it up) that Microsoft changed the traditional 'hand under a folder' icon for this feature so that if you looked carefully, the hand had it's middle finger extended.
R Shift + L Shift + Alt + Esc would trigger the debugger. I would use that to get the console back after btrieve/pervasive.sql ABENDed to unmount volumes before restarting, otherwise NetWare would spend an hour or two running VREPAIR.
My HS ran Novell in the 90s, and they learned this lesson after one of my friends (and fellow computer nerd) walked up to one of the 3 severs and typed DISABLE LOGIN (IIRC), thus locking out the math, English, and science teachers.
He spent the rest of the year disallowed from touching any computer in the school, which was interesting for our programming classes. I recall him making life for the librarians difficult by stopping by the library and making them look things up for him.
> Novell wouldn’t, or couldn’t, make Netware use task preemption and memory protection.
I actually remember being shown a pre-release version that had memory protection, at least. I don't recall if pre-emptive multitasking was included. It was a bit hacky; I think there was a kernel but then most of the OS ran in one process anyways, so perhaps it was a partial step. I think the problem was trying to be backwards compatible while shifting to the new memory model. Their development team was probably pretty small compared to something like Windows NT, so they didn't have the luxury of re-writing everything.
Looking at WikiPedia it is unclear if this upgrade was part of Netware 5 (released in 1998). Your other comment suggests maybe they killed this functionality. I was working for Microsoft by this point anyways. :-)
> There were some third party server applications built for Netware, I think maybe CC:mail server and Lotus Notes, but the memory protection and cooperative multitasking meant they didn’t last long.
We did actually have quite a few customers for the Oracle database on Netware. Mostly companies that were already locked into Netware for other reasons. In retrospect it does seem like madness to run an RDBMS on a server with no pre-emption or memory protection. I think the performance was pretty good, though. :-)
> Also Microsoft was at the height of its powers as a competition killing machine and nothing that stood in its way survived. It really genuinely looked back then like Microsoft was going to wipe out everything and it would be a Microsoft only future. Novell was no exception along with Lotus and Borland and many others.
This 100%. It felt absolutely inevitable that Windows NT was going to completely dominate this space once Microsoft came out with a competitive offering. And we didn't really know that Windows NT was a honest-to-goodness real operating system, instead of something hacked together on top of Windows for Workgroups (and DOS by extension). Of course, no one saw Linux coming at this point.
Not just around Netware but on the MacOS side as well.
The biggest issue for the entire industry at the time was the operating system. Everyone knew we needed real operating systems. Everyone was sick of crashing applications stomping on each other. Servers had just started to become a thing in the corporate world and server applications were just arriving. People were very interested in operating systems and where they were headed.
There wasn't even the concept of "bare metal" at the time. All DOS applications were bare metal. All Windows applications were cooperative multitasking and crashy. People were hungering for better.
Everyone expected that OS/2 was the server operating system that we would all be using. And indeed it probably would have been if Microsoft hadn't killed it. Lots of server applications got ported to OS/2 before Windows killed it stone dead.
Edit: for example, when I was in elementary school in the early 90s we had a lab of already antique Apple IIs, with no permanent storage or OS. We would just boot directly into programs on a floppy. Some of the programs themselves had simple features to save and open files.
Back when NetWare was being heavily used in the 1980s and early 1990s, 10Mbps ethernet was faster than accessing the local hard disk. ST506 (aka MFM) HDDs only have 5Mbps of bandwidth, and MS-DOS had no build in caching (until later releases -- maybe 5.x? My memory is fuzzy), but Novell NetWare ran in protected mode and had a real disk cache using all available memory on the server.
Once IDE HDDs became popular, the throughput of HDDs quickly began to exceed that of ethernet at the time. Once that happened, there was no benefit to centralizing storage resources on a file server when the local disk was far faster. It also didn't help that Windows NT was difficult to run over a network, but that was just another nail in the coffin.
I think that is a mistaken assumption. You have an effect, and you have a cause, but they are largely unrelated.
No, it didn't have preemption, or memory protection, early on, but that didn't matter because it was a file server and basically only a file server.
The Netware kernel was by and large a single binary running on a single core because that's all CPUs and PCs had then. It only did one job: reading stuff from data, sending it over the network, accepting stuff back and writing it.
It doesn't matter if you don't have memory protection if you only have one task. It doesn't matter if you can't preempt when there is only one thread.
When Netware was at its peak, it was the fastest fileserver there was, by quite a big margin, and that's all that mattered, because the apps executed on the client machines, not on the server.
I put in lots of Netware boxes and all they ran was Netware itself and nothing else. That's what it was for and what it was good at and all that I ever added anywhere was an on-server backup/restore tool, at the most.
Servers were not connected to the internet, and couldn't be, because Netware didn't include or use TCP/IP in the early days. They didn't route anything anywhere. There was nowhere to route it to or from, except other Netware servers on the same LAN.
They didn't have the killer feature you cite because there was no use for it.
I know that there were email servers -- that's where Pegasus got started. I know some people ran databases on it. I dealt with hundreds of clients from dozens of companies at 4-5 different employers and I never saw a single on-server app of any kind. No email, no databases, nothing.
It was a thing but not mainstream.
But when Windows NT came along, suddenly here was an OS that could be a stable file/print server and run apps as well.
You're saying that the world's best-selling bicycle lost out to motorcycles because it didn't have an engine. When it was the best-selling bicycle, motorcycles didn't exist yet. So it's true but it's not the cause. Motorbikes came along and allowed people to do a lot more stuff but bicycles still sell very well... without their engines.
What happened is that at Netware 4.x took over, Internet adoption spread at the same time. Suddenly ordinary small businesses wanted an email server, or a proxy server, and Novell did that -- for a price -- and they wanted it on the same expensive box.
And it wasn't very good at that. Then it became a problem, yes. But NT was doing well by then, it was a bit cheaper, client licenses were a bit cheaper, but more importantly, as others point out, NT didn't enforce client limitations whereas Netware did... and Netware was hard, while NT was easy.
So NT stole Netware's market out from under it.
Around the same time, the poor support for server apps became an issue. As did the difficulty of developing them.
Netware 4 let you push server apps out to a protection ring outside the kernel's Ring 0, but it cost performance. It did support some levels of protection but that was not the key thing.
NT was cheaper, easier, and while slower, ran server apps with aplomb. Even if Netware had been better at it, that wouldn't have saved it. It would still have been harder, meaning it required more skill of admins and developers... and it would still have been more expensive.
So, you're right, it was an issue -- but I think it was not actually a make-it-or-break it issue.
NDS made life hard for small businesses, who didn't want it or need it. An enterprise feature foisted onto SMEs.
And a new competitor, who was better in ways much more important than the ways it was weaker.
https://money.cnn.com/magazines/fortune/fortune_archive/1993...
> The NetWare operating system installed on the server computer acts as a traffic cop, masterminding the flow of information over a network. (The latest version costs $4,695 for 25 users, $47,995 for 1,000.)
NT was definitely cheaper, too. That made a difference in small deployments where I was often getting nickel-and-dimed.
Microsoft's license was soft, and depending on the customer buying enough paper to cover, if you could even find a reseller that could describe how to properly license Windows Server (exteremely doubtful).
OOTH, Novell's licensing was a hard limit. Once that # of workstations logged in, those without a license were hosed.
On all the jobs I did in those days, Novell networks at each customer site _never_ had enough licenses to cover every workstation. There was always somebody left out and they had to go bug people to see if they were done with the network so they could get in their tasks. Nobody ever seemed to overprovision their licenses to cover every single workstation (or even to have extra).
I never met a company using Microsoft server that properly had their CALs licensed, and they didn't want to hear it from me. Finding somebody to sell you the proper CALs was an interesting exercise.
Most other network server OSes at the time (the Unix vendors were trying hard in this space in those early days too) also had hard licenses. Which was part of why they didn't gain traction at all.
They could easily break the fake KMS but choose not to. It speaks a lot to their priorities.
Microsoft can tell us all they want that the telemetry is just for product usage. When the first slow quarter comes around they're going to gaze longingly at that huge treasure trove of data they have and start asking themselves if selling it won't help them meet their quarterly projections.
I'm unsure if this still works with Windows 11, but judging by how the upgrade process works, I would guess so.
- Bill Gates
Every small business was going to be running Windows already, so an NT server was kind of a logical conclusion. Couple that with the fact that the NT server was likely cheaper to license (even if it was ultimately not properly licensed), and that the NT server at least offered the appearance of being more flexible and capable than a Novell server, and it made sense that NetWare died off.
As a reseller, I also seem to recall needing to have a bunch of techs certified on all the various Novell apps to get a good discount. I remember going for my Groupwise certification specifically so our company could get a better discount level from Novell to try and be more competitive. I think I only ever installed one GroupWise deployment.
$4,695 in 1993 is worth $9,682 today.
$47,995 is worth $98,982 today
Because the biggest reason Netware lost was the inability to develop software on the platform.
This severely limited the value of Netware as a server operating system, as developers (for the most part) prefer to develop on the the platform (cross-compiling for other flavors of *nix doesn't count) for which they're developing software.
What's more, IIUC, the development tools for Netware were just awful. Which additionally limited the number of developers working on Netware-related applications.
And even those few folks mostly abandoned it as NT gained popularity as a server and a development platform. Had there been a robust user/development environment (like Unix), that likely wouldn't have happened.
[0] I am aware that Netware was mostly written in x86 assembler, but porting IPX/SPX and a few other bits and bobs would have allowed Novell to migrate to a Unix platform without having to rewrite the entire OS.
There was also a FOSS NetWare server for Linux at the same time too
A source, ironically from Novell docs.
https://www.novell.com/documentation/suse91/suselinux-adming...
It only emulated Netware 2/3, not the later NDS versions.
Uhh...The millions of clients who didn't have IP stacks?
I'd note that Novell didn't include an IP stack in Netware until 1996[0]. By then they'd already lost to Microsoft.
And they'd purchased Unix more than three years earlier, but didn't port Netware to Unix at all. Rather it was Netware on Linux, and wasn't released until 2005 -- when Netware was already irrelevant.
Had they migrated to Unix in the 1993-1994 time frame they would have eaten Microsoft's lunch. But they didn't. And so they lost. And they are gone.
If we are talking confusions that killed netware, my money is on marketing "network operating system" as opposed to "buy microsoft NT on your server to run your network"
Netware's demise was a matter of time. I am reading other comments here stating that it was the lack of preemptive multitasking that doomed it. Not even close. Netware was a feature mascarading as an OS.
I think Novell lost its focus and the other market factors also played a role. Novell's decline started when they also tried to compete with Microsoft for Wordprocessor market. Novell's then CEO became obsessed with beating Microsoft at everything . And also the Netware Directory services began losing to Active Directory as Windows 98 shipped with AD client built into it where as one had to manually install NDS clients on Windows. It was simply easier to manage your Windows deployments with AD than NDS. Moreover, IBM throwing in the towel with OS/2 ceded the desktop market to Microsoft Windows AD became the natural solution for managing and administering those desktops.
By the time Microsoft AD came around, there were actual uses for Microsoft AD, and plenty of buy-in.
The only thing Netware had to their advantage was a network of CNEs that made lots of money setting up and maintaining systems. But for small companies it was very expensive and large companies unnecessary as they had internal staffs that were not interested in another OS they had to train staff on. It also made them dependent upon 3rd parties to sell their systems and while it worked for awhile they in the long term will never be loyal.
The fact that the company lived so long on protocols ripped off from Xerox (they use to point people to Xerox documentation) is a mystery.
http://oldvcr.blogspot.com/2021/10/shiner-esb-apple-network-...
> Using a file server meant that PCs could share data with one another, for example permitting the first network-aware PC program of any kind – Novell’s network game SNIPES
Utter garbage. He didn't even say "first game" which would have been tougher. "first network-aware PC program" ?? get real.
> Novell’s proprietary IPX/SPX protocol
Which were Xerox's XNS protocols in all but name.
> TCP/IP – or Internet Protocol as we used to call it
No, we called it "TCP"
LAN Manager was a "joint effort" with 3Com (where I was). It was a stunningly bad deal for 3Com and we hated Microsoft at the time. It was Gates' first attempt to defeat Novell, and it failed at that.
All the same, Novell wiped the floor with 3Com in the 80s. Their stuff ran on every LAN hardware and ours, of course, only ran on Ethernet.
Re Snipes: I know of no earlier network-native app for the DOS PC.
It's described on Wikipedia as "Snipes [aka 'NSnipes' for 'Network Snipes'] is the first network application ever written for a commercial personal computer"
https://en.wikipedia.org/wiki/NetWare
A citation: https://web.archive.org/web/20160728221417/http://www.textmo...
If you have a pointer to anything earlier, I'd welcome it, but after all this time I doubt my editors will change the article. :-)
IPX/SPX was indeed based on XNS, as was Banyan's version, but it wasn't compatible with it. IPX was, I suspect, vastly more widely used than any other flavour of XNS ever.
> No, we called it "TCP"
Hey, it's a big planet, you know. You used 3 of the 5 letters, some others used the other 2 letters, or to make it easier, the words they stood for.
- http://www.mindcraft.com/whitepapers/nts4nw5filesvr.html
- http://web.archive.org/web/19990220215149/http://www.novell....
- http://www.mindcraft.com/whitepapers/response-novell-nts4nw5...
NetWare failed to innovate as an application server platform - NT had a much richer API.
NT was also viewed as much more easy to administer. This was a double edged sword as many people at the time didn’t (or wouldn’t) acknowledge it as a serious server platform due to having a GUI. Also it enabled many people to run servers who had no business running servers (tubes screensaver anyone?)
Finally Novell was really late grokking the internet. I remember running early NetWare 4.0x and trying to configure TCP/IP and having to configure IP addresses as dotted HEX quads. At that point I mentally flipped the bozo bit on them.
This is something I wrote in the early 2000s when they were making another (and pretty much last) attempt with SUSE. http://bitmasons.com.s3-website-us-east-1.amazonaws.com/pubs...
Everyone used Netware because they had to. Everybody hated it. I never heard a nice thing said about it.
When another, better, networking solution arrived, that was the end of Novell.
I learned to never invest in a company whose products people use only because they had to. They're doomed. Invest in companies where people love the products.
P.S. really bad investments are companies that like to sue their customers.
It's that simple. NT being an insecure, slow system that required a reboot every month didn't prevent it from taking over. And ten years later, Microsoft made most of those things better (but didn't fix them anywhere close to 100%)
The educational pricing for NetWare 4 wasn’t too bad and spent one summer planning and upgrading to it. Since it was small department it wasn’t too bad to administer, but could see the writing on the wall about the difficulties dealing with NDS clients on Windows and Mac workstations.
Don’t know what happened to the systems after graduation, but assume some of it was eventually consolidated due to security. All the desktops and servers had direct Internet connections and were not NAT’d nor had hardware firewall. That could not have lasted too many more years in the late 90’s.
Noorda made it go up, and he saw the beginnings of it going back down and that's when Schmidt was hired, probably in an attempt to turn the tide and avoid the inevitable victory of Microsoft over Novell. But Microsoft really had an unfair advantage and personally I do not think Novell ever had a realistic chance of staying in the race in the longer term.
He tried to bring it back but it was too late. Nothing he did could have mattered.