By the following I am only talking about a long-term solution.
Endpoint identifier /locator split : Yes definitely.
Each of these two items may get a structure. Lixia showed some for the
locator.
Wrt the EID: An AFI (address family identifier) may indicate the kind of
identifier. Currently E.164 addresses (telephone numbers) are mapped by DNS to
IP-addresses.
This is not necessary (at least in a long-term solution), if we have the
locator in addition.
There may be further address families we don't even know as of today
(e.g.one for all intelligent traffic lights:-)?). By going for multiple
address families (IPv4, IPv6 will of course also be such families) another dead
sin from the start becomes obvious: Different processing like unicast, versus
anycast, versus multicast, versus broadcast, versus mp2p, versus
mp2mp should be signalled by means of some mode of operation field in the
header - and not by separate address ranges. Imagine to have individual address
ranges for all address families and for all of these modes of operation
! What a bad design! Look at IPv6 : it introduces some anycast
address range instead of abolishing the multicast addresss range! This is my
view on EIDs.
My view on the split itself: See the gain for TE when traffic towards one
and the same egress locator can easily be identified for the sake of
TE-treatment !
A single word about the locator: This is the most
interesting object for all who are interested in future proven routing
technology, isn't it?
Heiner
In einer eMail vom 21.03.2007 23:28:40 Westeuropäische Normalzeit schreibt
tli@cisco.com:
On Mar 21, 2007, at 9:22 PM, Olivier Bonaventure wrote: |