Color Me Green, a New Experimental OS

The monster of modern desktop operating systems is apparently throwing money at the development of a Windows successor, code named Midori. This hyper connected internet OS built on managed code will apparently be able to run within Windows or a virtual machine, on x86, x64 or ARM architectures. A major focus of Midori will be on concurrency, based on an asynchronous only architecture. Parallel use of local and distributed resources with a component based and data driven application model.

The operating system apparently includes something called the Asynchronous Promise Architecture which will abstract details of physical machines and processors and behave consistently across local and distributed layers. Along side this will (naturally) be a programming model and platform stack. This will all be based upon services being provided across a trusted distributed environment. The programming model will assist developers to better utilize both local and remote resources.

Midori will post clear boundaries between processes to enhance inter process communication stability. It will enforce good coding on developers and aims to have strong notions of immutability. Multi master replication of complex data and variable wide variety power management features for a wide range of devices.

Beware! This is an incubation project, so when it fails, this may well become the next Cairo. Why would it fail?

Automated multi master replication of complex data isn't possible. Computers aren't decision making machines so can't attribute value to shared data. Is Joe User's document edit from New York more important than Sally Go Lucky's edit (on the same document) in London? Sure, an operating environment can identify when multiple edits have been made, and can even merge complex data ... as long as the edits are disparate. But when the edits are on the same data, human intervention (or automated data loss) will always be required.

Next port of call is yet another entirely new API. Is the 11th (or thereabouts) time the charm? There have been so many improvements to Microsoft's API since it's age old inception, yet applications still persist using the original API. This is an enormous thorn in Microsoft's side. The absolute success of Windows has meant that it continues to carry around baggage from 1986. It's this baggage, more than a lack of ideas, that holds Microsoft (and Windows) back. No business rewrites software that is fit for purpose, so even if Midori takes the world by storm (so far unheard of), the old API must survive and people will continue to use it for legacy software. Midori's programming model is starting to look like yet another API.

Simplification of the multi processing model for developers. As a systems software engineer with over 10 years experience in every commercial environment in use today, let me say that I can barely understand parts of Microsoft's Midori explanation let alone whatever it is that is highly unlikely to be released as a production development platform. Multi processing can be simplified, most definitely, and this is a good thing. The downside is that in order to provide this and be a serious asynchronous platform, the entire multi processing model must be both simple and fully accessible. It sounds like Microsoft has focused on creating something usable rather than workable.

Process isolation to maintain stability during inter process communication. Is Microsoft STILL trying to make a driver model that doesn't crash the operating system? Why are the problems in the Microsoft OS always caused by somebody else's poorly written code? Driver certification programs, managed code environments, user driver models ... urgh. What's the most insecure desktop environment in the world? Which operating system released in the last 10 years received the poorest user reception based on stability issues? What operating system has had the most listed security flaws in history?

Abstraction of process serving over a trusted network environment with applications written to make allowance for latency issues ... a fine idea for the future really. Think about what's going to be available in 10 years time. People think about modern network latency, but when fiber runs from one computer to another, latency will be measured as a heartbeat of what it currently is today. And fiber isn't that expensive, so it will only be a short while before it criss-crosses the globe like strands of hair. Here's the kicker for this issue. This is a Microsoft trusted network ... you know, like MSN was supposed to be or DCOM tunneling didn't quite manage or Microsoft domains never quite became or SOAP service integration is still struggling with. One big happy Microsoft network where everyone knows everyone and all are friends.

So the timeframe for Midori is some time after Windows 7, ~10 years. What this operating environment offers is a laundry list of solutions to problems we're facing today. In 10 or so years we should be far removed from these issues, otherwise we'll truly deserve Midori. But there's always hope. There's bound to be an ARM based net linked toaster or washing machine that will need an operating system.

Argghhh. Why do people eat up this 'news'? Why can't wealthy corporations bribe the news media? Oh, hang on ... ignore that last comment.