Toni -
I don't think that host identification is a necessary element of an
Internet architecture. It is important which software process you
reach, not which hardware this software process runs on. Load-balancing
mechanisms such as Akamai exploit just this. Similarly, after contact
establishment, it is important to reach the protocol state for the
particular session at hand, not which hardware this state is currently
residing on. Session mobility mechanisms exploit just this. So host
identification is not a general requirement.
Of course, individual Internet architecture solutions may use host
identifiers as part of service or session identification. Even the
existing Internet does this, as I mentioned in my previous email. In
those cases, the host identifiers are used in composition with other,
typically host-local identifiers, so that the composition identifies
either a service or a session.
So again, we need identifiers for only two purposes: for the purpose of
service identification, and for the purpose of session identification.
- Christian
On Jun 26, 2009, Toni Stoev wrote:
Christian,
There is one more primary need for identification on the general
network. That is the necessity of node identification. A node has to
have identifier in order to be distinguished from other nodes. Node
identification provides for efficient handling of node multihoming,
possibly dynamic, and for stable mapping of human readable names. One
basic use of node identification is connection identification, which
includes node identification. A complement to that use is the presence
of node identification in connection initiation. Node identification
descends from the domain name system and is currently woven into
unicast
addressing. Extracting it from both is the evolution.
Toni
_______________________________________________
rrg mailing list
[email protected]
http://www.irtf.org/mailman/listinfo/rrg
