Allowing a name to retain its values
-- other properties in the list --
allows it to be quoted with confidence.
Librarians and archivists require
references to hold over timescales
of hundreds of years.
Persistence depends not so much on
what is in a name, but on what is
not. The name must not depend on
irrelevant details whose value will
change with time, such as
- Internet host names
- Physical locations
- Dependedcies on organisations which
will change.
There is no reason why such details
should not be used to consruct a
name, and foir example to ensure
that it is unambiguous . The danger
is only when these details are necessary
for the reolution (if required) of
a name.
Message IDs
Therefore, persistence without resolution
is easy to achieve. Any hierarchical
scheme of names such as internet
host names, zip codes, etc, combined
with time stamps can be used as a
basis for generating unique IDs which
will not recur. The message identifiers
of Internet mail and usenet new are
examples of persistent names which
one cannot in general resolve to
the object itself.
Pseudorandom names
Another example is that a hash function
or digital signature can be used
to provide a pseudorandom number
which has an extremely high probability
of being unique. This method has
a special advantage of allowing anyone
to generate verify the name against
an original binary source of the
object.
Later we'll discuss ways of getting
persistence as well as resolvability.
When we look at current schemes,
for example in which aliases of internet
host names appear in object names,
we notice that the persistence of
a name is in fact limited not so
much by the technology, but by the
political and administrative issues
and human procedures which cause
domain names to come and go.
(up to requirements , on to unambiguity
)
Tim BL