In all cases of property rights there is a defined space, whether a namespace or physical space, and the task is to agree on simple attributes of or rights to control subdivisions of that space. In some cases a name or other symbol corresponds to a person or object owned or controlled by that person. For example, Internet users must agree on which domain name corresponds to which web site operator. In other cases we are simply concerned with control over a subdivision of the space. With real estate we must agree on who owns various rights (to occupy the surface, to mine the minerals under, etc.) to a piece of land. With radio spectrum we must agree on who owns what range of frequencies and in what physical space (or transmitting power as an easily observed approximation of physical space used).
...all such [multiparty problems of] control over the semantics of symbols, to be made and respected across trust boundaries, are problems of agreeing on and maintaining property rights...
...New advances in replicated database technology will give us the ability to securely maintain and transfer ownership for a wide variety of kinds of property, including not only land but chattels, securities, names, and addresses. This technology will give us public records which can "survive a nuclear war", along the lines of the original design goal of the Internet. While thugs can still take physical property by force, the continued existence of correct ownership records will remain a thorn in the side of usurping claimants...
The ideal title database would have the following properties:
(1) Current owner Alice should be able transfer her title to only a single relying counterparty (similar to the "double spending" problem in digital cash)
(2) Servers should not be able to forge transfers
(3) Servers should not be able to block transfers to or from politically incorrect parties.
...Using these results [of Byzantine quorum systems] it looks like we can approach our ideal title database as follows:
(1) Alice signs the title and Bob's public key, and sends this message to 2f+1 servers, committing her to transfer title to Bob. Bob checks at least 2f+1 servers before relying on Alice's transfer.
(2) No collusion of servers can forge Alice's signature (we achieve at least this property ideally!)
(3) A conspiracy of >=(1/4)n servers can block a transfer. Alice's recourse is to use some other channels to broadcast her intention, demonstrating that the registry did not follow her wishes, and hoping the alternative channels are more reliable. Bob only has similar recourse if he signed a document with Alice demonstrating their intentions to transfer title from Alice to Bob. The most basic recourse is a correct subset of servers which exits the property club and establishes a new one, then advertises its correctness (and proves the incorrectness of its rival group) as described above.