Gartner says: Windows is collapsing (really?)

If you are into IT, you probably heard the news that Gartner, the big USA research and advisory company recently warned the industry in a conference that Windows is in danger of collapsing. A bold statement, and once you read Gartner’s argument, it can give the impression to neophyte readers that it is a smart and throughfully researched argument. Well, not really.

As Paul Thurrott well described in his website, Gartner is just “playing catch-up” with the industry here. One thing I must agree with Thurrott is that analysts are “always late to the party”. But I would go further, claiming that analysts assertions are often driven by hidden agendas. A claim that, I must appologise to you, the reader, is way too broad to be covered here — and I don’t want to move away from the main scope.

It is not my intention to just second Thurrott in his opinions. In fact that would be a waste of time, as probably a good part of the IT community that is familiar with Windows and the recent strategies announced by Microsoft (particularly about Windows 7) is chuckling at Gartner’s comments. For those that would like to know why, you can read Thurrott’s views about Gartner’s assertions on his website here and then here.

What I do want in this case is to share you my views on what is the real problem with Windows. In my view, the biggest nonsense in Gartner’s argument is that they say Windows should have different kernels (“one size doesn’t fit all”) but at the same time they make a case against the over-complexity of the operating system.

Different kernels? This is as complex as it gets. My view is that Gartner simply wanted to sound smart and give the impression that they know what they’re talking about. As Thurrott well put in his site, Windows VISTA is the most modularised version of Windows yet, and there are several editions available for different needs (Home Basic, Home Premium, Business, Enterprise, Ultimate, …). The factual difference is that, instead of creating different kernels (as Gartner argues in favour) the key is to modularise it instead — a practice that is Microsoft’s main strategy for the Windows brand.

The problem I see here is that Microsoft should have opted for modularisation about five years ago. And I mean way more modularisation than Microsoft have now with Windows VISTA and Windows Server 2008. Things like the NTVDM, POSIX and OS/2 compatibility layers should have been optional modules. In fact, I was just recently thinking about NTVDM being removed from x64 flavours of Windows. Well, for me it makes sense. I mean, I can get much better DOS legacy support by getting something like DOSBOX to run DOS applications.

As for complexity, again I see this as a problem of modularisation. Just out of curiosity: What happened with the option of doing a “custom install” in Windows and being able to choose NOT to install Painter, Screensavers and things like that? For a 14GB of HD space hungry operating system like Windows VISTA, this is a must. I want to be able to choose what to add or remove during the installation, and without having to edit some unnatened script files.

Moreover, I think the complexity problem is way beyond bundled goodies like Windows Mail or Windows Movie Maker. The problem is that Windows VISTA requires a minimum of 1GB of RAM to barely run. I’m sorry but this is insane. I don’t care how cheaper memory is these days. What I see is that today you can do similar funky graphics in *nix with 512MB RAM. The problem? Again I think is the fact of little modularity within the kernel.

As for hardware compatibility, I agree with Thurrott that Windows VISTA being having issues of incompatibility is a myth. I know that VISTA has much, MUCH better drivers. However, this came with a dreadful cost (which is a related, but different subject):

The fact that Microsoft only allow signed drivers to run under Windows x64 is SUICIDE. Probably one of the siliest moves Microsoft did for the x64 platform. Why? Well, sometimes there are nice open source applications out there that require abstraction layers that sits on drivers (e.g.: network analysis tools, emulators, etc). Being open source drivers, it is well likely that they won’t have their drivers signed (me thinks). So, what if I want to install some open source cryptographic tool? what about VNC’s video driver? What about network analysis tools, and some *nix ported applications which require “drivers”? See my point?

Moreover, I despise the fact that Microsoft added some shenanigans on their HAL, such as disabling sound features when someone inputs something in the MIC IN or LINE IN of the PC, etc. These shenanigans, which were added to hurdle piracy, do little to prevent piracy but do a lot to annoy customers.

The way I perceive these issues, I don’t believe Windows is in danger of collapsing — far from it in fact. But I do think that Microsoft is in danger of loosing market within the enterprise and power user segments of the market to *nix (BSD, Linux, Unix-based systems) — And I make my point as an IT specialist and implementator, not as an analyst.