Microsoft's changing tune on Linux (News.com)

For example, several of the recent TCO comparisons have pointed to the fact that one person can administer a lot more Linux boxes than Windows boxes, and that's specifically because it's a ``command-line-focused approach'' rather than a GUI monstrosity. (Moreover, the Unix philosophy of ``let each tool do one job well,'' particularly as embodied in the multitudes of pipeline-capable utilities, adds even more bang to the sysadmin buck.)

Microsoft also continues to believe (or claim, anyway) that tight integration is a huge benefit to customers, but if you're running a bunch of servers, do you really want that GUI running all the time? Even if it's occasionally necessary, wouldn't you rather fire it up only when you need it, then kill it when you're finished so it doesn't continue to eat up RAM and virtual memory unnecessarily?

And then there's system replication--it's trivial to clone a pile of Linux disks from a single master (for example, with dd and tar), pop them into a set of vaguely similar machines, and have them all work from the get-go. But unless those machines are completely identical, good luck trying that with Windows. (Here I'm already assuming you can even get an appropriate license to do that from Microsoft. Not a problem with Linux and most Linux apps, of course.)

Even for non-sysadmin users, small home networks are becoming the norm, and it's hard to argue that Microsoft's lack of support for remote GUI applications (e.g., I fire up my office application on machine A but have it display on machine B) benefits anyone but Microsoft. Ditto for binary configuration files (death by registry, anyone?) and--ugh--drive letters that change when a new disk is added to the system.

Houston does make a valid point about consistency of the UI--that's always been a weakness of Linux and even Windows, to some extent (relative to Mac OS). But even though bringing that to Linux may require extensive work across numerous projects, fundamentally it's not an architectural problem and therefore can easily be addressed in an incremental fashion, with benefits that are progressively more visible to users. Windows' limitations, on the other hand, are architectural, and I think Microsoft is going to feel a lot more pain in the next few years because of it.

We don't compete in local economies for services business, so those dollars stay in that economy. We did a study and found that, on average, for every dollar that goes to Microsoft, nine dollars go to the local economy. And we think that's a very impressive ratio

In other words, "The total cost of deploying Windows is ten times as much as the software cost/licence fees alone". Is this figure consistent with the TCO surveys they report on?