[lars-b2018]

There is a good chance that something like what the author proposes is occurring, especially if you consider the Windows product distinct from the NT kernel. Does it make sense in the long run to support two separate Office code bases - the web based ones, and the desktop versions? I don't think it does, and I would think that Edge/Chromium is going to be their new base upon which the present day web based versions of office are going to be integrated into such that they can gain desktop access, similar to VS-Code and Teams running in Electron. VS-Code is wildly popular with developers, so there is some evidence that the user experience can be a positive one.

Large corporate applications are increasingly being built using web tech, which doesn't require Windows. If an organization can run a web based desktop and provide a remote desktop for "legacy" Windows apps, then this can reduce admin costs and improve security. The browser is increasingly the platform and the underlying OS just doesn't matter as much anymore.

At some point in the not too distant future, it won't make sense financially to continue to put a lot of R&D into the Windows kernel. It won't make sense financially to maintain two versions of Office and the huge amount of logistics that entails. If Microsoft can slide Windows over a different kernel (and it has been done before from Win95/DOS to NT), they can still preserve their user base, but reduce a lot of dev money spent on maintaining the NT kernel. Same goes for the legacy Office applications. If they can reduce their development spend, and focus resources on the app+win layer, this is what they are going to do.

[unilynx]

It seems unlikely that Microsoft could ever move Windows over to a different kernel and save on development costs that way, because no suitable kernel seems in sight.

For example the whole Handle and ACL infrastructure has no real equivalent in Linux (at most a bit on the filesystem level) and having to rewrite that for Linux seems pointless and would bring enormous backwards compatibility risks. Office is easy compared to the tightly integrated Windows Server/Active Directory stack.

But that still doesn't mean MS has to invest a lot of R&D in the NT kernel and stack. For userspace they indeed have electron, and on the kernel side they still have Hyper-V, and are even supplying patches to Linux for better integration.

Perhaps eventually Linux and NT will be mostly invoking virtual drivers with the real 'hardware control' work being done in the hypervisors, and the NT Kernel can just be considered 'done'.