(META: I've corrected the Subject to actually describe the
topic of this note.  I wish more people would do the
same -- so that Subject: lines reflect the actual topic
discussed by the note. :-)


1) As Joel and Tony have already noted, there is never a
requirement to use local-scope Identifier values.


2) Local-scope Identifier values exist only because there
are some (relatively small) user communities that desire
to use non-global Identifiers of various kinds.


3) The EUI-64 syntax is carefully designed to make it easy
to distinguish a local-scope Identifier from a global-scope
Identifier, so there is no protocol issue lurking here.



DISCUSSION:

Now, all that said, I don't see why this is suddenly
confusing to anyone.  All of the above is also true 
for IPv6 Interface Identifiers, which are also based 
on EUI-64, have some deployment.  Separately, IPv6
is purportedly widely understood and apparently has
some growing deployment.

In fact, the big difference between IPv6 and ILNPv6
(or ILNPv4 which also uses EUI-64 Identifiers) is that
the Identifier names the *interface* for IPv6, and
names the *node* for ILNP.  This semantic change is
absolutely crucial, because this change is what permits
IP sessions to remain up even though a correspondent
node's subnetwork point of attachment (SNPA) changes.

Clearly, node mobility is hugely simpler when a
global-scope ID value is used.  The ILNP specifications
say that normally a global-scope ID is used, while
allowing individuals that want to use some local-scope
ID to do so.  Again, no one is ever required to use
a local-scope ID value.

That noted, mobility can work fine with a local-scope 
ID, because the session Nonce permits differentiation 
of different nodes using the same (e.g. local-scope) 
ID values.  Implementations of ILNP are required to use 
cryptographically-random Nonce values, so an off-path 
attacker won't be able to predict a nonce value externally.

Existing IPv4 and IPv6, including Mobile IPv4/IPv6, 
are vulnerable to a wide range of on-path attacks 
unless IPsec is used to protect the IP session.  IPsec 
for ILNP works well even in the presence of changed 
node location, unlike currently deployed IPsec, and 
can be used to eliminate the threat of an on-path attacker.  
So there is no loss of security in moving from IP to ILNP,
or in moving from Mobile IPv* to ILNP mobility, or 
in moving from existing multi-homing methods to
ILNP multi-homing methods.

Yours,

Ran Atkinson


_______________________________________________
rrg mailing list
[email protected]
http://www.irtf.org/mailman/listinfo/rrg

Reply via email to