1. Does Codons tolerate Byzantine faults?  If a Codons server "goes
Byzantine" how much damage can it do?  What kinds of requests can it
cause to fail, and in what ways?  For this question, do not consider
faults in the underlying DHT: you may assume that the key-value store
itself always works.

2. DNS scales its hierarchical name space by delegating responsibility
for (nested) subtrees of the namespace to subdomain servers.  In
general, name operation is handled by the DNS server for the smallest
subtree containing the name.  Ceph uses a similar structure to scale
its metadata service.

(a) What happens if responsibility for a subdomain moves to a
different DNS server, e.g., due to an administrator action or failure
of a server?  Can this event cause an interruption of service for
lookups in that subdomain?  If so, how bad is the problem.  Why is
this design choice OK, i.e., what would be the cost to avoid any
interruption?  Consider this question for Ceph as well: what if a
subtree of the name space moves to a different server, e.g., due to
MDS rebalancing?

(b) Codons improves scalability of the namespace by hashing and
replicating name records across the collection of servers.  Would a
similar approach work for Ceph?  If not, what is different?  What are
the tradeoffs?