In all honesty, my understanding of ethereum is probably insufficient. From what I understand, ethereum is a blockchain application, similar in a number of ways to bitcoin, but instead of simply changing values in cells (send 0.5 bitcoin to X address), ethereum transactions can execute code (which are called contracts), and cells have persistent memory which can be altered in numerous ways by contracts. It costs ether to execute these contracts, which is a way to determine if the execution of the contract is complete, and prevents network-bogging loops. Theoretically, applications could be built within ethereum, allowing for the decentralization and public validation of the application itself. Thus, a voting application could run in ethereum, and it would be transparent to all involved.

Obviously, one concern about ethereum's practicality is the potential size of that growing blockchain ledger.

The development news I linked addressed some of the kinds of operations that might be performed in a contract, how the state of an address is altered or read by the operations, and about lowering the overall memory involved in some operations.