The Stanford Archival Vault (SAV) is a digital object archive that
reliably stores objects for long periods of time (decades or
centuries). In order to retain and use snapshots of the Internet,
objects should be indexed and replicated to protect against
failures. We discuss techniques for indexing objects when they cannot
be deleted or modified. It may also be useful to select subsets of the
Internet (e.g., certain collections of websites) for replication. We
present a reliability algorithm which describes how to define and
replicate such subsets so that they grow implicitly when a new related
object is added to the archive. We also examine the challenges faced
when building a useful user interface for the system.