Inside Windows 7 -- what we know so far

You’d be forgiven for missing it, but Vista turned a year old last month. November 30, 2006, marked the official launch of Vista to the business community: the first part of an odd two-stage liftoff which was followed two months later, on January 30 2007, by the mainstream consumer launch at which Vista became available to one and all.

(Unfortunately for Microsoft, the number of people who bought Vista was much closer to ‘one’ than ‘all’. The majority of ‘sales’ of Vista have been from Microsoft to its OEM partners for pre-loading on new systems, and even then several companies chose to offer buyers a choice between XP and Vista rather than foist the shiny new OS onto the public. Just because they bought a new PC didn’t mean they wanted to ditch their familiar old OS, especially when they had all the drivers and software and everything worked pretty well on the whole, thanks all the same.)

So now the clock is ticking on Windows 7.0, the successor to Vista (which was officially Windows 6.0, following Microsoft’s lead of counting iterations of the NT-based kernel rather than actual client editions).

We’re still in the long dark before 7’s dawn, but the earliest signs are encouraging: a new streamlined kernel, an inbuilt VM for running old software, a revised and simplified UI... there’s every chance that Microsoft intends Windows 7 to rise from the ashes of Vista and be what Mac OS X was for Apple.'

is this Windows 7: this screenshot, floating around on online forums, purports to be from an alpha of Windows 7. Probably fake, but interesting nonetheless.

No name

For starters, you can forget about its initial codename of Vienna. Newly-minted Windows veep Steven Sinofsky has nixed the fancy cyphers set by former Windows head Jim Allchin, which were to represent cities with wonderful views or ‘vistas’. Of course, the final name that appears on the box will be anyone’s guess, especially after such unexpected flights of fancy as Me, XP and Vista.

But we’ve got a hunch that Sinofsky might return to the simple year-based branding which Microsoft introduced with Windows 95 and has since applied to much of its OS and client software. After all, it’s done well enough for Office, which was under Sinofsky’s rule during the era from Office 2000 to Office 2007 (as well as heading cross-suite work for Office 95 and 97).

...but maybe a number?

If that’s the case, Windows 7 might end up as Windows 2010, because that’s how far away the new Windows will be. Microsoft has flip-flopped on the cadence of its desktop client over the years, sometimes promising a new OS every 18 months (usually after the arrival of an OS which was years overdue), othertimes two years or more.

The official word on Windows 7, however, came in July during Microsoft’s Global Exchange sales conference in Orlando, when a spokesman said that “Microsoft is scoping Windows ‘7’ development to a three-year timeframe, and then the specific release date will ultimately be determined by meeting the quality bar.”

So that’s at least three years from Vista’s arrival, or two years from today, and maybe then some – which means late 2009 through to early 2010 as the initial timeframe, but marching onwards through 2010 if need be. Okay, so maybe it’ll be Windows 2011...

Or is this it: another supposed screenshot of Windows Seven from osbeta.org...

Virtual machines for ‘legacy’ software

There have also been indications that Windows 7 will use virtualisation to run any software that hasn’t been specifically written for Windows 7 or using Microsoft’s .NET language. 7‘s use of virtual machines to run these ‘legacy’ applications was leaked on Microsoft’s own Channel 9 community forum in July in a (quickly removed!) thread.

While it's a novel approach for Microsoft to take, it's certainly not a first. Apple migrated users from its Mac OS Classic environment to Mac OS X by loading the classic OS in a virtual machine of sorts if users needed to run one of their old applications.

At this early stage, no-one can guarantee that any feature will definitely be on the Windows 7 roster. After all, at the equivalent stage of Vista’s evolution Microsoft was talking about everything from the WinFS database storage system to all manner of ‘blue sky’ notions, all of which were dropped before Longhorn hit its first beta release.

Sinofsky’s track record paints him as more of a realist, however, and OS-based virtualisation makes sense for plenty of reasons. Microsoft already has the technology in Hyper-V, the hypervisor-based virtualisation system designed for Windows Server 2008.

And hardware won’t be an issue: by the time Windows 7 arrives circa 2010, quad-core will have replaced dual-core as the mainstream, with substantially larger cache including big slabs of Level 3 cache.

L3 already exists in AMD’s ‘Barcelona’ architecture and have been hinted for Intel’s ‘Nehalem’, which will succeed the current Core micro-architecture in the second half of 2008. (In fact, if Windows 7 breaks cover towards the end of 2010 it’ll be accompanied by Intel’s post-Nehalem Core microarchitecture revision, codenamed Gesher.)

Also, considering that Nehalem will debut with eight cores in a single die, there’s no reason we couldn’t see a string of single cores each being set aside for running a VM, with a flash drive used to hold and launch the virtual machine software in order to dramatically boost session speed, especially during the ‘transition states’ of startup and shutdown which represent so much of the VM overhead.

PCs will also sport obscene amounts of memory: 4GB will likely be equivalent to today’s ‘entry level’ of 1GB, with flash drives used in concert with hard drives to actively store files rather than just be a shot-term cache.

And one more possible Windows Seven screenshot: This one looks rather like a doctored version of Vista, but you never know...

A line in the sand..?

It’s worth noting that virtualisation is comparable to how Mac OS X handles Mac OS 9 software: a ‘classic environment’ is launched, creating a sandbox instance of OS 9 (although you can run only one Classic process per user).

Apple used this approach to make the leap from OS 9 to OS X, which was in reality an all-new operating system sporting a UI which mimicked the more familiar elements of OS 9.

Might Microsoft be planning the same seismic shift for Windows? The OS itself has already entered in its second decade, and the 32-bit NT codebase underpinning the current Windows generation is already nudging 15 years from its 1993 launch in Windows NT 3.1.

There’s no reason to rule out the concept of Microsoft resetting the clock to zero with Windows 7: releasing an all-new OS built from the ground up as an operating system for 2010 but with the ability to support pre-7 ‘classic’ apps – which by then will mainly be relatively modern XP and Vista software – in virtual machine sessions.

Kernel knowledge

Eye candy, begone: MinWin is so lean that even the Windows flag on the splash screen is rendered using ASCII

We do know that the next generation of Windows will be built around a stripped-back ‘microkernel’ codenamed MinWin. As previously reported, MinWin has been described as “the Windows 7 source-code base”.

MinWin is currently an internal project to strip back the NT kernel to the barest of bare metal, but will be used “to build all the products based on Windows” said Microsoft engineer Eric Traut during a demonstration of Microsoft’s virtualisation technology at the University of Illinois in October.

“It’s not just the OS that’s running on many laptops in this room, it’s also the OS used for media centres, for servers, for small embedded devices.”

As ‘proof of concept’, Traut showed an iteration of MinWin consisting of just 100 system files, which occupied 25MB of hard disk space and ran in 40MB of RAM.

“It’s still bigger than I’d like it to be, but we’ve taken a shot at really stripping out all of the layers above and making sure that we had a clean architectural layer there”.

The return of WinFS?

More speculative is the question of WinFS, which sits atop the NTFS file system to allow data to be stored, accessed and managed based on relationships with other data. WinFS was originally to use the Yukon database engine of SQL Server 2005, which included native support for XML, but became the first of Vista’s many ‘foundation pillars’ to topple -- primarily because Microsoft couldn't get the speed of the system remotely close to something a user would consider acceptable compared to the relatively simpler NTFS file system in use today.

Despite initial promises that it would be released in the year following the launch of Vista, the last news on WinFS was that some of its technologies have been rolled into the Katmai engine of SQL Server 2008. Microsoft may well forge ahead with a relationship-savvy file system in Windows 7, built around the Katmai engine, but the ‘WinFS’ label could remain buried.

Red, as in hell: another possible screenshot of Windows 7?

A new look

There’s no doubt that Windows 7.0 will sport a revised interface. It was Sinofky’s winning gamble to give Office 2007 an all-new UI which swapped the decades-old clutter of menus, toolbars, task panes and what-not for a single task-aware ‘ribbon’.

Office 2007’s UI overhaul itself was led by Julie Larson-Green, who (as reported earlier this year) Sikofsky has since tapped to head the “User Experience” program for Windows 7.

Will Windows 7 see as radical a facelift as Office 2007? That’s harder to tell, because the change in Office 2007 wasn’t made for change’s sake: Larsen-Green went back to first principles for the suite, and she’s likely to do exactly the same for Windows. Starting with a clean slate, she’ll be asking what people expect their computer to do, and then how an OS should fit in with that. But its safe to say that feral UI elements such as Vista’s ‘icon overload’ Control Panel are not long for this world.

Windows to go fluid: the designer of the Office 2007 "fluid" ribbon interface is in charge of interface design for the next version of Windows

Roll your own?

The nuked Channel 9 post which placed virtualisation on the Windows 7 blueprint also indicated that the look of the UI would be highly customisable, perhaps implying a de-coupling of the top-most user interface layer from the actual Explorer shell.

(It’s wor