If you are or know a student in the UK, you'll know about Midori. It's a painfully sweet clear green concoction with the great taste of melons, somewhere on the other side of the galaxy from a decent Islay malt.
Midori is also the code name for a non-Windows operating system under development at Microsoft. As yet, few details exist outside the company - but there's a very interesting overview of what's known over at SD Times, which claims to have seen some internal design documents relating to the OS.
There's a lot to get past the tonsils, although many of the ideas will be familiar to anyone who's thought about the current challenges in system design: concurrency, parallelism, service oriented architecture, heterogeneous platform support, managed code, replication, and resource management. In all of these areas, Windows is very weak -- and all of these areas are becoming increasingly important.
There's not space in this blog to cover the implications of all of the above, let alone explore the real problems of migrating an industry even assuming Midori comes out more Laphroaig than lager top.
A couple of details may give the flavour, though. One of the underlying technologies being discussed is the Bartok compiler and runtime system. Currently a product of the Search group, Bartok examines the code it generates for known classes of error - which seems a most intriguing use of ideas that have grown out of search and anti-malware research. This may be the start of practical self-analytic systems that have their own sense of correctness and the ability to check against that at all stages during their lifetime, and that has implications way beyond catching potential stack overflows.
The other tiny detail that stood out was resource management. This is a low-level part of Midori that understands details of the capabilities of the platforms it's running over, and makes intelligent decisions about scheduling and resource allocation based on such ideas as processor, memory and power available. At one level, this brings to mind the concept of launching a particular application or service from a mobile device, with the operating system finding the right place on the network to run the meatier parts while leaving the mobile coping with those bits its good at.
But at a deeper level - and I've no idea whether this is part of the game plan - it leads to the idea that active management of energy consumption can be wired into the system to very fine detail, with much more finesse than the current "Is anything happening? Right, let's go to sleep" processor-level control. Most IT burns most of its power doing nothing much; anything that truly addresses this has world-changing implications.
But all this is in the future. I hope Microsoft indulges in its better nature and opens up the design of Midori for community discussion and work. Trying to distill an entirely new computational infrastructure for the world is hard enough: trying to do it in the dark is just plain crazy. It's bound to come out the wrong colour.