You seem to have made a leap in reading my note that I did not intend.
At least in the model I was attempting to describe, the stack ID is NOT
used by the routing system for reachability, and not normally used by
the packet forwarding system for forwarding.
(If one has a design in which the stack ID appears in eveyr packet, and
if one has other constraints that lead one to want to, one might choose
to do last hope resolution using the stack ID instead of the locator.
But that is, usually, a detail that can be worked out. It does affect
what locators need to name.)
Yours,
Joel
Dae Young KIM wrote:
On Fri, Dec 4, 2009 at 1:57 AM, Joel M. Halpern <[email protected]
<mailto:[email protected]>> wrote:
As for what I want to identify, I tend to think that the identifier
should correspond not to the physical node, but to the logical stack
construct. A logical set of protocol machinery, stacked from the
network layer up. A physical machine may have multiple of these.
And in some cases, a single one of these may be operating across
physical boundaries, if the coupling works right.
This is a model I could understand. In old terms, which many of this
community might not like, this 'logical stack' id is once used to be
call 'node address'. And this exactly 'node address' is used to identify
that logical stack construct (from network layer up through transport up
to the app layer) AND, what is more important, is also used for
'routing'. Stating again, one id, called 'node address' is used both for
identifying a node AND routing.
So, in this model, there's no two separate things like ID and Locator.
Whereas... the 'physical address', or in Internet terms 'PoA address',
is in fact the Layer 2 address, or MAC address.
So, you
- identify a (logical and physical) node by a 'node address'
- do you L3 routing based on this 'node address'
- in passing through sub-network or L2 network, you map this
'node address' to 'L2 or MAC' address.
The flaw that the Internet had was to regard the IP address as the PoA
address. No. IP address should be the 'node address'. That is, if you
have to choose one of the two roles that IP address was thought to
assume, i.e. ID and Locator, it should be the 'ID' part we have to
preserve. The name 'ID', though, is not the correct one, since in this
(new but in fact old) model, this 'ID' is also used as L2 routing.
Therefore, the correct name should be 'node address'.
Then what to do with the PoA part of IP address we gave up above? That
PoA part will exactly be the role of sub-network or L2 (or MAC) address.
So, to me, IP address is fine as it is now, only that I use IP address
as pure L3 'node address'. Or put it more correctly, 'sub-network
independent internetworking' node address.
Whereas the physical address you mentioned is to me, either
- subnetwork address or
- L2 address or
- MAC address
depending on what underlying service is available to my
'internetworking' sublayer.
The whole confusion of the Internet has ever started to define IP
address as PoA address. No, it should be 'node address'. And there's no
such thing as PoA address, but subnetwork or L2 address.
I don't propose to throw away IP address. Rather keep it but use it only
in the context of the 'node address' (which term this community has
associated with ID).
Mostly, I like your idea of separation. Only that, your ID should not be
a new creation, but the IP address be used as ID, or more exactly, node
address. And leave everything else to the underlying layer's job or
property.
--
Regards,
DY
http://cnu.kr/~dykim
_______________________________________________
rrg mailing list
[email protected]
http://www.irtf.org/mailman/listinfo/rrg
