As IT Week revealed recently, Microsoft is working to make Windows run Linux software, and not before time. Windows and Linux would be much more useful platforms if they could each run software originally written for the other.

Now let's talk about Longhorn. Longhorn is the codename for the next version of Windows, currently slated for delivery in 2008. Though Microsoft has mentioned the possibility of a 3D graphical user interface, and a new database-like file-system, it has actually made no promises about what will be included in Longhorn.

Many people note that 2008 is so far away that without an excellent crystal ball, nobody could predict what the update will contain. But one thing I reckon for certain about Longhorn is that it will be able to run Linux software.

Such a move would provide Microsoft with some excellent ammunition to win business from companies looking to deploy Linux, and those looking to replace Unix-based Risc systems with Unix based-x86 ones. If Microsoft wants to sell more server software, adding this capability to Windows would be a fine way of doing it.

In fact, the new "server roles" installation technique due for delivery with Longhorn is perhaps the only other certainty about Longhorn. This is Microsoft's new approach to software deployment. Rather than install everything all at once, the server roles approach allows IT staff to tell the installation program exactly what a particular server will be doing, so the installer only installs the software needed for that task.

IT Week identified the drawbacks of Windows installing unnecessary software last year in our review of Samba 3, which outperformed Windows as a file server in our tests. Part of Microsoft's explanation was that Samba and Linux combined require far less resources than Windows.

I talked this over with the guys in Microsoft HQ, and I'm sure I could hear the tears hitting their desk as they realised that an operating system so bloated with optional extras was doomed to lose ground against products designed for a single purpose.

In instances like this it is clear that less really is more.

Anyhow, server roles would also provide an easy way to install Windows Services for Unix (SFU). This is important because in its current form, SFU is quite tricky to install. For example, you either need a working NIS server or a few files containing user names and encrypted passwords to be present on your c:\ drive. These things are easy enough for Linux administrators to arrange, but they can form an impossible barrier to Windows administrators attempting their first installation of SFU.

A server roles approach would reduce the resources required by an SFU system, but more to the point it would provide a way for customers to build Microsoft Unix servers that did not need to be weighed down by insecure Windows software, such as Internet Explorer or other unnecessary Windows tools. Imagine that.