The VVM is our answer to several problems. Here are just a couple.
It is intended to be the ultimate environment for supporting agent-based interoperability, in both a distributed sense (agents move around between platforms) and a local sense (arriving agents can be written in different bytecoded languages, and can expect to share data with one another seamlessly).
Replace "agent" in the above with "application" and you arrive at our second goal: an extensible, dynamically-configurable, execution platform for arbitrary bytecoded languages.
Of course, the space between ``agent'' and ``application'' is immense. Some of the more interesting occupants of The VVM Interzone are smart cards and embedded systems: a very high-level VM implementation language reduces the time to market for smart-card and similar technologies, and aggressive support for ``soft'' VMs permits painless upgrading of embedded systems ``in the field''.
The open research questions raised by the VVM cover a remarkable range of fascinating disciplines: implementation and optimisation techniques, interpreter and compiler theory, language-neutral data representations, semantics of programming languages, security and behavioural verification, and extensible operating systems -- to name but a few.
Don't expect too much yet: we have only just started on this, and our resources are severely limited.