I've mentioned about virtualization at the OS level and even at applications install/run level but true virtualization need to be taken by the individual applications such that both OS changes and app version changes will not affect the data itself.
What I mean is that data is almost always looked upon as belong to a specific version of software today. If you want to work with new format, you have to upgrade the software. That won't work if the OS becomes virtualized and the app versions differ between the two versions of OS. Data need to survive the version changes: too often apps support "backwards compatibility" by taking in old format and converting into new format but that shouldn't be. Customers don't care what version the data is. They want to work with their data and then pick and choose the app version and the OS version [sometimes you are forced to use certain version of app to work with specific OS version, but that's app problem, not user problem].
Hopefully the trends towards xml and OS independent apps [like Java apps and web based tools like google map] will open up this world of data independence.
Labels: virtualization