Meanwhile, in the real world, hundreds of millions of users of Windows desktops have no admin access to their systems. Very clearly the OS works perfectly capable without that being needed.

As for Android and Chrome Books, they abandoned trying to pretend to be Windows and have been successful as a result (well, Android anyway).

Linux will be far more compelling when it supports business systems, when it supports gamers, when it supports creatives and when, as a result of at least two of those three, it starts to become the default installed operating system on mass market personal computers.

Oh, and yes, I've been using Linux for longer than you've been evangelising it for desktop.

The Steam Deck (or more accurately, Valve's Proton, which enhances the Wine project) seems to be the final push necessary to get gaming on linux as a thing. So that's one off your list.

> Meanwhile, in the real world, hundreds of millions of users of Windows desktops have no admin access to their systems. Very clearly the OS works perfectly capable without that being needed.

Because end-users should NEVER have admin access, that is why.

[Linux will be far more compelling when it supports business systems]

Red Hat and others have been doing okay for decades selling Linux for business.

[when it supports gamers,]

Steam Os and Steam Deck. Or just install Steam on a compatible Linux distro. You have STEAM, and can run a lot of Windows games on Linux thanks to the custom Wine Steam uses.

[when it supports creatives]

Linux does supports creatives, the main problem is the comercial apps being Windows or Apple only. Blender and Gimp run on Linux and have improved a whole lot over time, they are just harder to use.

Blender is so confusing it is hilarious but yes you can still use it. Gimp is slightly less confusing but I don't use it much so it is always confusing for me.

Recently Dreamworks is saying they will Open Source MoonRay so of course there will be a Linux version.

[when, as a result of at least two of those three, it starts to become the default installed operating system on mass market personal computers.]

Microsoft literally lobbies and sometimes pays money to have Windows preinstalled in every machine they can.

The biggest problem Linux has nowadays is that there is way too many distros, that is great because it gives you a choice but it also does confuse people.

Heck I actively use three different Linux distros but most people prefers one size fits all.

For that I say, use Linux Mint, it can do everything, maybe not perfectly but it can and it is more or less user friendly.

-> Because people use those IPCs to get work done.

This is the reason why businesses (and most people) run Windows - the choice and availability of apps. There are not always direct (or as good as) equivalents of Windows software on Linux. LibreOffice is not as good as MS Office. LibreOffice Base is about 20 years behind the times, for example, and that's if you can get it to run at all without crashing - I've tried several times and just gave up trying to get the right combination of OS + Java. Then there are all the adds ons and tools that are available for MS Office. A new LibreOffice Impress theme doesn't quite have the same value. I write this with regret about LibreOffice as I use it, and I don't mean it to be churlish towards the LibreOffice project.

Gimp is not an equivalent to Photoshop, if you are in the design world. And the Gimp GUI appears to be designed by a greybeard Perl programmer.

Part of suggesting that people use certain software, whether it is free or paid for, is being honest about it. While businesses can in some cases use Linux on the desktop, in others they simply cannot. This bald statement that "Businesses should dump Windows for the Linux desktop" does not do Linux any good at all. I can imagine the response in numerous cases, it would be along the lines of "we tried Linux and it (meaning the apps they need) didn't work or were not available".

Maybe that's a good thing?

20 years ago, office was fine, then they added loads of cruft we don't need.

In any case, Micro$haft is pushing us into the cloud where office365 is online. So that stuff is getting less important anyway.

These days, Access is about 10 years behind the times. When did it last get a meaningful update?

I'm pretty sure Microsoft now wants you to use PowerBI instead of Access. Other, usually server-based alternatives may be more appropriate for your needs.

Excel, I think is where the biggest problem is. Libre Calc will do the basics, but it does not have feature parity with Excel.

Access is probably 20 years behind the times; I doubt it does much more now than it did in 2006 when I used it to do a major project that depended on some of its unusual, and powerful, capabilities. But it was hitting its limits then, being a 32-bit program with small (by today's standards) file size limits. On the other hand, Libre Office Base looks like an undergraduate project from 1977, capable of doing the tiny "wine list" demo but not any real work. Never did, and probably never will.

Excel is a special case. The 2010 version was great. Later ones have been unstable; the code base must be a stinking mess. A special place in hell for the fact that it often loses the ability to receive input from the keyboard! (Workaround: Mouse over to a different spreadsheet, type into it, then go back to the one you're working on.) So I often find myself falling back to an old installed version with a warning on to that it is no longer supported, but at least it runs.

Yes, WIndows 10 Pro, not Home, and I don't feel too insecure. I just take normal precautions, like not opening spam attachments.

Access is probably 20 years behind the times; I doubt it does much more now than it did in 2006 when I used it to do a major project that depended on some of its unusual, and powerful, capabilities. But it was hitting its limits then, being a 32-bit program with small (by today's standards) file size limits. On the other hand, Libre Office Base looks like an undergraduate project from 1977, capable of doing the tiny "wine list" demo but not any real work. Never did, and probably never will.

Excel is a special case. The 2010 version was great. Later ones have been unstable; the code base must be a stinking mess. A special place in hell for the fact that it often loses the ability to receive input from the keyboard! (Workaround: Mouse over to a different spreadsheet, type into it, then go back to the one you're working on.) So I often find myself falling back to an old installed version with a warning on to that it is no longer supported, but at least it runs.

Yes, Windows 10 Pro, not Home, and I don't feel too insecure. I just take normal precautions, like not opening spam attachments. The OS has many flaws, but it is designed for people to use, not for programmers to use. And as Three Dead Trolls In A Baggie sang over two decades ago, Every OS Sucks.

True, but how much do you really use in MSOffice ?

I would say that LibreOffice has 99.9% of what everyone needs to actually work.

It's only manglement that needs those pretty charts to make themselves feel like they are important.

And if management want you to send them the data which they can easily handle, without going through hoops to convert or munge it, what then?

99.9% is probably far too high a figure if you want to interact with other people/companies.

At work, the top feature I use of office is communal editing.

I have several word documents and excel documents. A team of four (or more) people are actively, throughout the day, editing these documents. We can't be emailing them back and forth, we need our edits to reflect immediately.

We need to be able to see a log of who changed what. We need good visibility of proposed changes and comments.

All of that, we have with the current generation of MS Office with the O365 plugins.

Are there alternate ways of doing some of this? Of course. But the technical barrier of entry is very low. I don't have to teach the less technical members of the team some tool to make things work or collect the data needed for status updates.

Never underestimate the power of an effective visualization. I spend a lot of time in tools like Splunk figuring out the right visualization to most effectively communicate to non-Splunk experts.

Sure, only 20% of Office features are used by everybody, but everyone uses perhaps a further 20% of its features and (the important bit) it's a different 20% for every user. I guarantee that every single one of the other 80% of features is used by somebody, and if you're in a large organisation and threaten to take any of them away, you'll be deluged by "but I need that feature to do X which is essential for my work".

It isn't as simple as you seem to think.

Some 20 years ago I was teaching CS at a local college and was managing a laboratory with Linux machines for the programming classes. The computers had just the basics: a text editor, one free IDE, browsers, etc.

The lab worked great for our purpose.

Some people in a local branch of the government got wind of this and decided to copy the idea to save money on Windows licenses. After they converted one of their labs their boss came to visit and asked about the software they planned to use, and I got a frantic call asking "where can we download Autocad for Linux". They got really upset when they learned that they could not run all their applications on Linux as if it were a free, compatible version of Windows.

(icon for the "20 years ago" bit)

Is there an equivalent on Linux for stuff like SCCM (Config Manager) so admins can build machines to a corporate build with zero touch, and can roll out software remotely easily. Every time I've looked at Linux this seems to be a major hole and when you are looking after 100+ sites it is impossible to send technicians out to install software for every user, and 75-90% of our users are not capable of doing installs themselves.

If there is something now, I'd be interested in having a look at it again.

I agree with your sentiment. I design and build full IT solutions to support industrial automation activities as an OEM. Love it or hate it when you have 500 Windows servers/workstations you have a single Active Directory domain to administer. When you have 500 Linux boxes you have 500 Linux boxes to administer. I want to use Linux but I need the enterprise management tools that Microsoft brings to the party and are sadly lacking in the Linux space. If Linux (from any vendor) had an equivalent to GPOs I'd jump onboard.

I agree with what you and AC above wrote. Alas what you will get with Linux is "you can try Puppet or CFEngine or...". While those tools do work, they are not the same thing as AD. Horses for courses. I have no issue with those being used for server farms where each machine is more or less identical and usually running one application (e.g. web server). But that's a different prospect from the desktop.

You might want to check out https://www.freeipa.org/. It covers a lot of AD functionality. You can even set up a trust relationship with AD.

To be honest, I never looked at things like Group Policy with it as we used Ansible on the project. That's not a bad thing, just different. With things like PXE, Ansible-pull, and/or Ansible Tower (upstream AWX) you can create images that configure a box from scratch without much intervention past the boot stage.

I know I'm not convincing anyone here, but it is possible and frankly (to me) does it in a more understandable and logical way than any of the (very dated) AD methods I've tried.

Windows Admin for 20+ years now, that uses linux for my desktop.

FreeIPA, god awful thing compared to AD.

Windows deployment via WDS and Group Policies is far easier to get going and working then using things like packer, ansible, setting up a PXE server etc.

Along with keeping windows configured via GPO, far easier than ansible with something like tower/awx.

This is coming from someone that has migrated from AD GPO to using ansible, paker, terraform, tinkerbell, and maybe adding awx.

The advantage of going with something like ansible is that you can use it to configure you windows and linux hosts, plus your network, your OOB controllers, gluing different systems and automation together. Its more work if just for your hosts, but does give you more flexibility later on to integrate with everything else.

For example, currently finishing some ansible roles and playbook to deploy esxi hosts, by providing only the mac and service tag of the IDRAC. Finds it on the network, sets the vlan, sets the DHCP reservation, configures the iDRAC, certs, AD integration, etc, setups the RAID config, Configures the zoning of the FC fabric, deploys ESXi, configures esxi and joins it to vcenter and puts it in the correct cluster.

Well, all the tools are there, automated installations, both scripted and from images has been around on linux for more than 20 years, Active Directory works nicely with linux for those who want that, I you rather do your central management by some other tool, NIS, LDAP among others has also been around for a long time.

SALT, Puppet, Chef are all just a few examples the range of mature solutions that has been around for a long time. Several popular distros are providing their own tools for those who does not fear vendor lock in; Landscape, YAST etc.

What actually may be lacking, is what Windows admins call "Best Practices", popular and well proven solutions that most admins approve of. The abundance of tools paired with easy scripting of most everything makes Linux system administration a much more diverse field. Google, Facebook, Alibaba, Twitter, Amazon, Ebay, Wikipedia and all the other large players would not even consider touching linux, if they could not automate their large scale deployments. The fact that those companies are running far more linux servers than linux desktops, does not alter the level of truth in my conclusions above. On the other hand you will naturally find more information about using the tools stated above for server management, but the will work equally well for the management of desktop computers.

You'll get something similar to "best practice" guides, if you opt for a RHEL-based distribution (Rocky, AlmaLinux, Oracle Linux, etc.). The Red Hat documentation has a lot of examples of good practice along with many how-to guides.

It's ironic that the very foundation of AD is a number of tools and protocols that came out of work done on UNIX and other UNIX-like OSes. I'm talking here about Kerberos, NTP, LDAP and no doubt a number of other technologies.

All three of the things I've explicitly quoted have been subject to EEE, such that now most organizations use MS software to drive them. And they then put barriers to make it more difficult to integrate UNIX and Linux systems into the environment (for example, withdrawing the GUI tools for managing UNIX user authentication around the time of Windows server 2016). You can still do it, but it's a bit of a masochistic exercise and uses techniques mostly foreign to the majority of MS administrators (like firing up what are effectively text editors to manage the maps).

What the two commenters above have pointed out is that, unlike Microsoft environments, there is no one single way of doing what they want in Linux. Top down tools work in Windows environments, because if MS makes a tool, it is extremely likely to be deployed almost everywhere (before AD, there were a number of Windows herders around, all of which were assigned to the bin when MS moved into that space).

Even if you include Red hat, Canonical, SuSE et. al. no one organization has the clout in the Linux space to make a single, one-size-fits-all configuration and management suite. There are plenty of tools around, many widely used in niche (or not so niche) environments, and some that pre-date AD, but nothing that is all-encompassing to the point where you would never consider using anything else, and there probably never will be.

This is again the advantage and curse of Linux. There's choice.

I'm pretty certain that if proprietary UNIX had remained around for longer (especially if it included desktop systems, as it used to do), that eventually the likes of Sun, IBM, HP, and DEC would have come together, and defined a single management tool that worked across platform, and standardized it in an official standard, like they did with Motif, CDE, or DCE. But nothing like that is ever going to happen in Linux space, because it's just too easy to fork and do your own thing without buying into the expense of the standards game.

We only use Linux laptops in our IT department, each laptop is pre-configured using an Ansible script by someone in IT, and then sent out to the user. Part of that is installing OCS Inventory, which monitors the laptop for its state and stores the results centrally, and also allows pushing out and configuring new software.

As someone who has always run and managed my own OS on my workplace computer, I didn't like this loss of control and having someone backdoor'd into my machine, but as you say, its an essential part of maintaining a big fleet of laptops.

Isn't that just called a distro in linux? Your own corporate distro can be as simple as Debian with different default settings or a few extra default programs.

SCCM has a few moving parts. Rather than a single Linuux product being a complete replacement, you might consider one or more of the following:

Rudder - configuration compliance: https://www.rudder.io/

Ansible - a shallower learning curve than Chef or Puppet, I think, and doesn't require an agent: https://ansible.com

Rundeck - task automation: https://www.rundeck.com/

Foreman - lifecycle and patch management: https://www.theforeman.org/

FreeIPA - identity and policy management: https://www.freeipa.org/

These are all open source, so you can use them FoC. Commercial power-ups exist.

I manage several large systems and hundreds of desktops for Scientists and Engineers. I insist on systems hardware being a similar as possible. Once I have a system cominality. I use Tar and grip through a script to clone the base system. Then as new systems come in I extract the tarball onto the new machine. Change the host name if necessary. Otherwise that is controlled through scripts. If DHCP then it is determined by the MAC address. My favorite are the diskless systems.

Everyone gets the same programs. The Scientists are constantly conducting seminars on their unique applications. The problems I get involved with are environment issues. Where someone wiĺ cludge grep, gawk or some other internal Linux utility. Because they have a long held dependency on a system setting that they expect everyone else to live by. So I et the engineers and scientists decide how they want that piece of code to behave. Ibhave learned a lot from these lady's and gentlemen. The biggest save for me is Environment Variables. A utilIty by LANL. Which allows one to alter their environment on the fly by loading and purging an environment for their work. No need to custom tune each user's environment anymore. That was a huge save for everyone. For a unit of 300 scientists and Engineers. We have one administrator me. Hardware problems a seldom. All machines have ECC memory. So pinpointing a defective DIMM is a piece of cake. Hard drive failures are easily recovered due to bare metal backups again using tar and gzip.

Trying to do this with Windows would not be possible. I even have three students from the local University who are studying how I manage my systems. Their findings are impressing the University and helping them to create a course of study on Linux systems administration.

My biggest headaches come from those who just hate Linux. Want it gone. The problem with that is Windows slows down productivity. Which is the key focus for our organization.

"another problem with Windows' single-user ancestry is that the default Windows user all too often must run as the all-powerful PC administrator"

I've been managing Windows networks in education since 1999 and I've never come across anything that's needed admin access to run.

And there's no greater hacker out there looking to exploit the system than a bored schoolboy.

Yeah, this whole article was written by someone who clearly last used windows around 1994 and seems to think that nothing has changed since then.

"..."another problem with Windows' single-user ancestry is that the default Windows user all too often must run as the all-powerful PC administrator".."

I suspect by this what you mean is that by default the first time a new Windows system is set up, the user details you use will be given local administrator rights.

However - this is only true of a home build. Every corporate using tools will not allow this and even with the zero-touch configuration of Intune, you have the option to not let the user be added to the local admins group.

The comment above about Excel being the biggest problem is spot on. Let's face it, here we are in 2022 and businesses across the world still run on Excel spreadsheets. We know that is a problem but it isn't going away any time soon.

Add to that, so many of the big software companies will do things like create Excel or Outlook plugins which only compound that issue.

You can avoid the whole online Office route right now by installing Office 2019 but expect that option to go away as soon as MS can make it disappear.

Applications, tools and plugins are the biggest reasons you won't see many places suddenly say "let's go to Linux as our main desktop solution".

Some evidence to support that statement would be nice.

Nevertheless, while I'm no advocate of Windoze and (as a security professional) have no respect for its supposed 'security', the major barriers to transitioning to Linux desktop seem to me to be the extent of the current MS user base and incompatibilities between notionally comparable applications for the different OSs (e.g. MS Office versus Libre/Open Office). If documents break when transferred between systems, there's a strong disincentive to make the transition as statistically speaking almost all business documents are still generated using MS Office.

If this problem could be fixed I'd move to Linux in a flash, but for now I'm stuck with MS as my customers demand Office format documents.

My experience is that Ms Office (well, Word) breaks formatting between machines in the same damn company. So there's that. We are also averse to moving everything to Azure, as some stuff would not be allowed to. Then, Microsoft is now actively disabling stuff on-prem, since all and everything have to be in the cloud. Which sucks. A lot. Let's see if this provides us with the needed push.

On the other hand, most of our production machines / servers are Linux (SLES? I think....), apart from the stuff I have to work with (stupid SSRS etc.). So why not the desktop as well? I would also gladly trade in the MS SQL server for something else. Only problem is, I have a few colleagues who are learning resistent...

let me think of what I use every day, that would be problematic on Linux.... apart from the tSQL centric stuff? Nothing. LibreOffice is plenty good for what most of us do. Oh, but I have to admit that maybe outlook might be hard to replace. The whole integration of calendars, group mall boxes, etc. is pretty good (though it crashes too often on my machine). But then I have not looked too much into alternatives (evolution?).

It's more than a little out of date -- about 15 years ago there was still a lot of software that worked this way, but practically nothing off the shelf needs admin rights now, and the problem is usually bespoke software.

Either way -- any business that still has software that needs admin rights to use is going to be something that IT have already tried to get the business to replace, and have failed. It'll be that same software that would wreck a Linux migration.

If only some large body had thought of this! And tried it! Oh wait, Munich did, and gave up. Scheslwig-Holstein is having a go, and it's going to take four years ish, if it succeeds. I work in local Govt, and there's no way we could completely abandon Windows, as we get involved in so many different things, and use niche apps that we just couldn't afford to get redeveloped for Linux. We'd only be able to get rid of on prem Windows by running Windows VMs in the cloud for the niche stuff, and that would get expensive.

Meanwhile, many years ago when I worked in Academia, the Dean of the Faculty told us to 'get rid of Windows',,... now, I was the Dept Unix sys admin, but I was rather concerned for my colleagues job, who looked after the Windows servers and desktops,... so the first person I migrated to Linux was the Dean's PA. By lunchtime she was crying in the cantine and threatening to quit if she didn't get her Windows PC back. That ended that. Going Linux is fine in theory, for small groups of technically minded people, but for a large organisation that has any measure of staff churn? Training will be an issue, you cannot expect new hires to be competent with your chosen flavour of desktop / software. Then factor in sending and receiviing data and files from partner organisations who still use Windows, we have close ties to the NHS, Police and education sectors, all of whom are Windows users. Oh, and we provide bookable computing resources to the general public in our Libraries. Jo Bloggs probably can steer Windows OK, and possibly wants to polish his CV in Word. Sorry Jo, you have to learn a new Office Suite, but at least you can put that on your CV,....

As I recall, Munich reported > 90% success. Yes, the niche stuff had to retain MS. Then an incoming (fanboi) mayor demanded his Outlook or Nothing, so all that valuable work had to be rolled back.

Dumping a new system on somebody's desk and providing no training or support is just indefensibly vicious and unprofessional. Incompetent if I'm being generous, although your description makes it deliberate sabotage. Not Cool. (And don't forget that the next version of Word will need people being retrained because they'll have shuffled the commands again. )

BTW, when I worked in support I often had to help Win/Office users adjust to OS X and Appleworks. (Yes, really. It easily met our needs.) Show them around the basics, be prepared to look up questions I'd not heard of, maybe buy a few copies of "The Missing Manual." Job jobbed.

Wow, aren't you massively judgemental? I had no budget fior training. I didn't use Linux Office Products, so didn't know them myself, to teach to others, and training staff wasn't my job. We had an internal training team, who also did not know Linux or Linux Office Apps. I did exactly as I was instructed by the Dean with the resources he allocated to me (literally nothing).

As to niche stuff, that's the spoiler. Local govt will need to keep some Windows, so how then do they integrate with their peers? Only use Windows for the niche stuff? That makes keeping Windows massively expensive, and sites will need to allocate both Linux and Windows resources to staff. It would have been interestign to see how Munich would have managed during the pandemic, if they'd carried on. We just started using Teams.

Yep, what's your point?

Really I don't get why virtualisation hasn't already solved this problem. It's always been possible to remote host niche apps and give users really basic workstations... or even dumb terminals.

Then you can reserve the lion's share of your silicon for the CAD users, rather than leaving most of your potential CPU cycles idle on the desk of desk jockeys who use Word, Excel and a browser...

Virtual Desktops is just really expensive to implement. You need a hypevisor environment (possibly Nutanix, so you've introduced a new OS into your support model) You still end up managing the thin client hardware (so have to support the base OS on the thin client, set up a management server / cloud management) and then the virtual OSs you intent to use, so Windows and Linux. You've gone from possibly a pure windows estate, to managing four operating systems. You still waste CPU, because sizing the CPU power of the hypervisor is a black art, and even though there are tools to measure the usage of your clients and give you a number for your cluster,... if you get that just right first go I'll give you a medal.

@Greybearded old scrote "Dumping a new system on somebody's desk and providing no training or support is just indefensibly vicious and unprofessional. Incompetent if I'm being generous, although your description makes it deliberate sabotage. Not Cool."

I also read it as deliberate sabotage, but necessary instead of "Not Cool". It looks like the Dean of the Faculty expected Linux to be rolled out without any training. Rolling it out to the PA meant the Dean got the full impact of the rollout before it had hit the rest of the Faculty.

It could have gone 2 ways switch back to windows or provide training before continuing the rollout. Looks like the Dean to the easy option.

My recollection is much the same as yours re Munich.

f the only way to convince the Dean that getting rid of Windows and replace all applications can't be done without preparation and training is to reduce the PA to tears something is very wrong.

Either the Dean is completely immune to feedback or the one tasked with advising him on IT matters is missing critical social skills (or a combination of both).

Pointing out that both the unix admins and their colleagues supporting the windows desktops and applications would need training to even be able to plan (not to talk about roll out and support) a migration properly and estimating costs and time frame would probably have stopped the Dean's idea as quickly and without collateral damage to the PA.

I although do not think there would be less support required after the migration; unless the former windows administrators are unwilling to learn a new skill set they would still have a job.

The Dean was a tyrant and a very angry man. Sometimes the only way to educate such people is to give them exactly what they ask for. He left and joined the clergy.

I'd have done the Dean first, unless he was Linux evangelist.

Your recollection is incomplete. 90% was the peak figure for user desktop replacements, but very quickly those users began to request a return to Windows as the Linux-based solutions were missing key functions.

The Linux ecosystem has no centre, and its continuous development model makes it much more of a moving target for internal application developers than Windows is, unless you expend a fair amount of effort maintaining your own distro. (Package maintainers are the real heroes of Linux - it’s a thankless, time-consuming job, but without them the whole ecosystem would collapse).

The nail in the coffin was the prospect of decentralised working. There simply wasn’t any good private-cloud solution for collaborative work. For several classes of application, Munich basically came to the realisation that to make Linux work for it, it would have to permantently invest in its own team of programmers and maintainers in addition to its existing IT staff (who also had their own grievances regarding provisioning and monitoring). It was a buy/build decision, and the real costs of “build” were now known to be far higher than simply paying for a ready-made product. After all, the city doesn’t build and maintain its own trams - the latest lot was bought from Siemens - so what made software any different?

Other organisations have fared better with Linux, however, but they were more pragmatic: adopting it where it was a viable and mature alternative. The mistake Munich made was the (ultimately ideological) decision to do a top-to-bottom replacement of every software system with a FOSS alternative. That really makes no sense - you might not like Exchange, but is there really a FOSS alternative that’s anywhere near as good at managing office communications?

Yes, and 90% just isn't good enough. As far as the users are concerned you are providing degraded functionality for a nebulous and theoretical advantage. I worked for years to keep microsoft out of my organisations server estate for what I believed were good reasons, but it was a losing struggle. More and more niche but vital applications were an increasing struggle to work with, with vendors who had neither clue nor interest in an environment that wasn't 100% Microsoft. And in a big organisation with 4 figures worth of applications the struggle was never ending. I managed to take early retirement before they finally elected to move over, but I have heard rumours that it was a far more difficult process than was imagined.

-> As I recall, Munich reported > 90% success.

And there is the problem. 90% is not good enough. Going from a (presumed) 100% success on Windows to 90% on Linux does not make the cut. It is nowhere near good enough.

If after half a day an office worker is crying in the canteen and threatening to quit about having to get to grips with the fairly minor diffferences between Linux and Windows its not the OS that needs binning.

How would she have managed when Windows 7 HAD to be replaced with Windows 10?

> Training will be an issue, you cannot expect new hires to be competent with your chosen flavour of desktop / software.

That goes double for whatever abomination to choose to handle your timesheets/holidays/docs repository etc. Windows > Linux is the very least of it.

"Given a choice between security and having applications work as expected, Microsoft often chooses the insecure status quo"

I suspect a lot of businesses (at least those where the IT department exists primarily to provide a service to its users), would feel that having applications work as expected is a fairly high priority. After all, there's plenty of things you can do to mitigate against security problems, but without apps that work, your business is dead in the water.

One of the reasons Windows is popular is because it gave developers the ability to hack together these department-level apps using access and word macros and the like. Trying to solve the problem by cutting it off is like saying "cars driving fast are dangerous, let's limit them to walking pace"

FYI I hate Windows and use Linux for my day-to-day, but wouldn't recommend it to anyone who isn't an experienced IT geek.

How many Linux boxes I saw badly configured because its primitive permission system does not allow granular permissions - not everything is a simple web server or database server. And people do try shortcuts under Linux as well.