Where is the responsibility on the companies and organizations that want to run this ancient and insecure software on such mission critical applications? Microsoft does offer a path for these organizations that can't upgrade to a newer OS, they charge an onerous support fee. This is intentional to make it painful for these companies to continue using a product that Microsoft knows is insecure. They are trying to incent good behavior (aka upgrade), like the government does with a tax on alcohol and cigarettes.

If you are in a highly regulated environment like the UK NHS then there is no excuse for either not being current, paying the proper fee Microsoft to support the OS you choose to continue to run, or taking other measures to ensure that your systems are protected, such as keeping them on an isolated / secure network with no Internet connectivity. We have solutions for this stuff, Microsoft isn't the bad guy here. The people that consciously made the budgetary decision to disregard their customer's / patient's data / welfare are responsible for this.

I'm no Microsoft fanboy, but blaming Microsoft for this is like blaming Ford for a traffic death that occurs today in a car that was manufactured in the 1950s before seat belts were standard equipment. We now know seat belts save lives, if you chose to take the risk of driving a car without them that's on you, not Ford.

Seat belts can be retrofitted, and most car schematics are well known. Windows cannot be modified.

This is more like blaming Ford for a road accident caused by faulty brakes on old cars that they knew about and didn't recall.

There was a Microsoft Research project that made it possible to run Windows XP (or Server 2003, I don't remember exactly) as a PV guest.

That was really cool, but the whole project disappeared.

If it had been open source, I bet it would still be actively maintained to this day.

Edit: Found the paper https://www.microsoft.com/en-us/research/wp-content/uploads/...

I will admit I've just rapidly paged through that PDF, but it looks like I'm reading a Xen introductory paper.

Xen is open source.

I found some PV IO drivers at https://wiki.xen.org/wiki/Xen_Windows_GplPv which mention XP (search for 'XP' including (!) single quotes), and a quick Google does immediately give hits on running XP as a HVM guest.

I'm (genuinely) curious what you're describing/referring to here. What project disappeared?

He's talking about Windows on Xen, which existed at a time, but was never released, like a lot of research projects. AMD-V and Intel VT made it mostly moot though.

So you mean like... NTOSKRNL et al essentially retrofitted to run in a kind of userspace?

Nice.

I don't expect that kind of thing to ever leave a research environment though. It would mess with too many people's heads and give people too many ideas of running bare-metal kernels other than NT.

Now I think about it, I realize the reason why HW virtualization really took off is because it let vendors keep their operating systems as actual operating systems in the traditional sense of the word, making for fewer legal issues (among many other reasons).

Also, I thought Xen was essentially just a super-thin layer to kickstart VT-x/AMD-V. I didn't know it could do anything else. In fact, I thought there was only emulation and hardware-assisted virtualization. Is there a middle ground I'm not aware of?

Yes. It's paravirtualization. Oh, Drawbridge is full NT in user-space, is in production now for SQL Server on Linux, but Dk (Drawbridge) is much newer. :)

TIL. That's really cool. Now I'm wondering if there are any small fully-paravirtualized hypervisors and guests I can play with. I guess Linux's support for various forms of I/O acceleration is more or less it.

I didn't know Drawbridge was that amazing - that's incredible.

And now I'm starting to understand Microsoft's vision: they have WSL to get Linux infrastructure onto Windows, and Dk to get selected Windows infrastructure onto Linux. Impressive.

But now I think about it that way, I know Dk will only ever be an internal framework - if that got released we'd basically have "perfect Wine" and it would allow quite a few too many applications to move off of NT.

WSL uses Drawbridge picoprocesses internally by the way. :)

The Drawbridge NTUM(User-Mode NT kernel) is maintained as NT 6.2 (Windows 8) which is new enough for almost all purposes - except modern Windows apps.

There's also Drawbridge...

Someone the other day wrote an excellent comment explaining why you can't just replace Windows with Linux in some professional environments like hospitals: medical hardware drivers that are only available for Windows.

I'll try to find and link that comment that managed to make a better point than I did.

You can unplug the ethernet cable, however. Or at least firewall it.

This is what we did 10 years ago when I worked in a hospital.

We knew these devices were insecure by default. Some even shipped with a network enabled MS SQL Server with a blank sa-password. Quite literally a free root-kit.

Scientists and doctors working on these machines were forced to use portable storage (floppies, ZIP-drives or CD-RWs).

It was cumbersome, but no network was a strict policy, and it was there for a reason.

That's refreshing to hear that whoever had the authority in that situation also had a brain.

I wonder how tricky it would have been to set up a MAC- and plug-location-based VLAN to isolate those devices onto, with a very very carefully locked down machine sitting between the devices and the rest of the network. Deep packet inspecting firewall, copious logging, antivirus turned up to 11, the works.

I ask because I'm curious how well a theoretical setup like the above would have worked out for the described scenario - I'm sure there are similar environments where it may be impossible to get having no network approved by management.

Do we even know medical devices were affected, the reporting has been so shallow I'm not sure whether they have or it was just administration.