Two main schools of thought guide the adoption of BYOD in the enterprise. One is to reduce the risk of the devices themselves by managing them closely through policy and software. The other is to reduce the risk to data that may be exposed or lost through mobile devices.
The latter is happening by default in most companies while everyone considers whether the former can be done. All organizations need to decide on which overarching BYOD strategy will guide all their individual BYOD projects -- or agree that different pockets will use different strategies (which usually isn't optimal).
[ Also on InfoWorld: Data security in a BYOD world | Understand how to both manage and benefit from the consumerization of IT trend with InfoWorld's "Consumerization Digital Spotlight" PDF special report. ]
The latter strategy, focusing on securing the data, separates the device from the data -- which can be accomplished in several ways. Many different solutions are being developed, including using Web services, virtual machines, virtual desktop integration, and virtual application integration. Most of the "unmanaged BYOD" vendor offerings focus on one of these types of solutions.
I believe this focus on data security is the best strategy for many reasons, not the least of which is that keeping unmanaged devices off your network would stifle productivity. BYOD is inherently unmanaged, and in trying to control it, you'll always be putting a square peg in a round hole.
At the same time, you don't want end-users connecting to highly sensitive data via systems that are at major risk of being compromised, without any offsetting controls. That would be foolish.
This basic idea behind this data security strategy dates back to the 1960s and is known generally as the "red/green paradigm." As you might expect, the green part is for low-risk systems, while the red system is supposed to be used for all high-risk operations. The two are logically or physically separated -- but unfortunately, every previous attempt at this sort of differentiation has failed.