Client-driven data duplication / redundancy

[gavin-norman-sociomantic]

We may enhance the dhtnode to support data duplication / redundancy, but it's also worth noting the old idea that the client could be enhanced to perform this.

Something like:

• For particular channels, a level of duplication is configured.
• When writing to a duplicated channel, the client would actually assign multiple Put requests to different nodes, based on the level of duplication desired.
• The keys of the duplicate records could be generated by some algorithmic means, e.g. re-hashing.
• When reading from a duplicated channel, the client would try to contact the responsible nodes in order, until the record is retrieved.