TO: v6ops working group
Here is DJ's review.
Two more experts are in the progress of
reviewing this deliverable on behalf of the
16ng working group. It will be available soon.
Hope this helps.
Daniel (Soohong Daniel Park)
Mobile Convergence Laboratory, SAMSUNG Electronics.
======================================================
My thoughts...
2.1. Last sentence of first paragraph..
This paragraph reads like 802.16 is some sort of competitor to 802.11
through is wider footprint and improved mobility. In reality they are
highly complementary technologies. I suggest to change..
"So it is expected that IEEE 802.16 network could be the next step of
IEEE 802.11 network."
To
" So it is expected that IEEE 802.16 network could be the next step ofin
IEEE 802.11 wireless networks.
2.1 Second bullet point
The wording below needs clarification.
"
o BS: Base Station. A generalized equipment set providing
management and control of MS connections. There is a
unidirectional mapping between BS and MS medium access control
(MAC) peers for the purpose of transporting a service flow's
traffic. A connection is identified by a connection identifier
(CID). All traffic is carried on the connection. Sometimes there
can be alternative IEEE 802.16 network deployment where a BS is
integrated with an access router, composing one box in view of
implementation.
"
Change to:
"
o BS: Base Station. A generalized equipment set providing
management and control of MS-BS connections. There is a
unidirectional A transport connection is a unidirectional mapping
between BS and MS medium access control
(MAC) peers for the purpose of transporting a service flow's
traffic. A connection is identified by a connection identifier
(CID). All IP data traffic is carried on the transport connections.
Sometimes there
can be alternative IEEE 802.16 network deployments where a BS is
integrated with an access router, composing comprising a one box
in view of
implementation.
"
2.1, the diagram.
The following diagram describes far more than an 802.16e network that
the text claims:
“
Figure 1 illustrates the key elements of IEEE 802.16(e) networks.
Customer | Access Provider | Service Provider
Premise | | (Backend Network)
+-----+ +----+ +----+ +--------+
| MSs |--(802.16)--| BS |-----| | | Edge | ISP
+-----+ +----+ | AR |---| Router |==>Network
+--| | | (ER) |
| +----+ +--------+
+-----+ +----+ | | +------+
| MSs |--(802.16)--| BS |--+ +--|AAA |
+-----+ +----+ |Server|
+------+
Figure 1: Key Elements of IEEE 802.16(e) Networks
“
You might change the text to say:
“Figure 1 illustrates the key elements of IEEE 802.16(e)WiMax networks.”
Or
“Figure 1 illustrates the key elements of IEEE 802.16(e) networks
typical mobile 802.16 deployments.”
“
1. Lacking of Facility for IPv6 Native Multicasting
IEEE 802.16 PMP mode is a connection oriented technology without bi-
directional native multicast support.
“
Well this is true, but it is also true for modern Ethernet. The
difference is that in the Ethernet model (it defines a point to point
link), the presence of an 802.1D relay is assumed to be present, but the
same assumptions are not being made for 802.16. 802.16 connections
joined by an 802.1D relay do support native bidirectional multicast.
So two cases exist..
BS joins connections together with a bridge => bidi mcast ok
BS doesn’t => we have a collection of mostly independent tubes.
I suggest
“
1. Lacking of Facility for IPv6 Native Multicasting
IEEE 802.16 PMP mode is a connection oriented technology without bi-
directional native multicast support. IPv6 neighbor discovery
[RFC2461] supports various functions for the interaction between
nodes attached on the same subnet, such as on-link determination and
address resolution. It is designed with no dependence on a specific
link layer technology, but requires that the link layer technology
support native multicast. To support bidirectional native multicast
and 802.16 BS must deploy additional mechanisms, such as 802.1D
bridging. The absence of this This lacking of facility for IPv6 native
multicast results in inappropriateness to apply the standard neighbor
discover protocol specially regarding, address resolution, router
discovery, stateless auto-configuration and duplicated address
detection.
“
“
2. Impact on IPv6 Subnet Model
IEEE 802.16 is different from existing wireless access technologies
such as IEEE 802.11 or 3G, and, while IEEE 802.16 defines the
encapsulation of an IP datagram in an IEEE 802.16 MAC payload, a
complete description of IPv6 operation is not present. IEEE 802.16
can rather benefit from IETF input and specification to support IPv6
operation. Especially, BS should look at the classifiers and decide
where to send the packet, since IEEE 802.16 connection always ends at
BS, while IPv6 connection terminates at a default router. This
operation and limitation may be dependent on the given subnet model
[I-D.madanapalli-16ng-subnet-model-analysis].
“
A CS looks at classifiers and decides not where to send the packet, but
how. I.E. on which CID. That CID would be one of the set of CIDs that
links to the same peer CS. The difference is in the QoS properties that
apply to each CID, not the denstination.
So the classification in the CS is rather moot. I suggest
2. Impact on IPv6 Subnet Model
IEEE 802.16 is different from existing wireless access technologies
such as IEEE 802.11 or 3G, and, while IEEE 802.16 defines the
encapsulation of an IP datagram in an IEEE 802.16 MAC payload, a
complete description of IPv6 operation is not present. IEEE 802.16
can rather benefit from IETF input and specification to support IPv6
operation. Especially, BS should look at the classifiers and decide
where to send the packet, sSince one end of an IEEE 802.16 connection
always ends at
BS, while an IPv6 connection terminates at a default router. This
operation and limitation may be dependent on the given subnet model
[I-D.madanapalli-16ng-subnet-model-analysis].
2.2.1.3
“
Note that in this scenario IPv6 CS may be more appropriate than
Ethernet CS to transport IPv6 packets, since there are some overhead
of Ethernet CS (e.g., Ethernet header) under mobile access
environments .
“
For fairly static Ethernet headers (like a link terminating to IP at
both ends, PHS eliminates most of the overhead. I suggest…
Note that in this scenario IPv6 CS may be more appropriate than
Ethernet CS to transport IPv6 packets, since there are some overhead
of Ethernet CS (e.g., Ethernet header) under mobile access
environments . However PHS, if deployed, mitigates much of this overhead.
2.2.2.1
“
The BRAS in Figure 4 is providing the functionality of the AR. The
Ethernet bridge is necessary for protecting the BRAS from 802.16 link
layer peculiarities. The Ethernet bridge relays all traffic received
through BS to its network side port(s) connected to BRAS. Any
traffic received from BRAS is relayed to appropriate BS. Since
802.16 MAC layer has no native support for multicast (and broadcast)
in the uplink direction, the Ethernet bridge will emulate Ethernet
level multicast (and broadcast) by relaying the multicast frame
received from the MS to all of its ports. Ethernet bridge may also
provide some IPv6 specific functions to increase link efficiency of
the 802.16 radio link (see Section 2.2.2.3).
“
This text assumes that native multicast is a property of Ethernet. I
used to be for fat-yellow-coax, but isn’t for modern point to point
ethenet. The bridge does not emulate Ethernet level multicast. In the
802 architecture, the bridge is the thing that implements multicast, not
the medium (although exceptions exist).
I suggest:
“
The BRAS in Figure 4 is providing the functionality of the AR. The
Ethernet bridge is necessary for protecting the BRAS from 802.16 link
layer peculiarities. The Ethernet bridge relays all traffic received
through BS to its network side port(s) connected to BRAS. Any
traffic received from BRAS is relayed to appropriate BS. Since
802.16 MAC layer has no native support for multicast (and broadcast)
in the uplink direction, the Ethernet bridge will implement emulate
Ethernet
level multicast (and broadcast) by relaying the multicast frame
received from the MS to all of its ports. The Ethernet bridge may also
provide some IPv6 specific functions to increase link efficiency of
the 802.16 radio link (see Section 2.2.2.3).
“
“
2.2.2.3. IPv6 Transport
Note that in this scenario Ethernet CS may be more appropriate than
IPv6 CS to transport IPv6 packets, since the scenario need to support
plain Ethernet connectivity end-to-end. However, the IPv6 CS can
also be supported. Every MS and the BS has the Ethernet type MAC
address. If the MS is using IP CS, then the BS needs to take care of
the Ethernet header. In the upstream direction, the BS will need to
generate an appropriate Ethernet header and prepend it to the IP
datagram. In the downstream direction, the BS will use the
destination address from the Ethernet header to identify the MS and
then it will strip the Ethernet header before relaying the IP
datagram over the 802.16 MAC connection. Ethernet bridge may provide
implementation of authoritative address cache and Relay DAD.
Authoritative address cache is a mapping between the IPv6 address and
the MAC addresses of all attached MSs.
“
This description does not seem to fit with the definition of the CSs.
When using the IP specific part of the Packet CS, there is no Ethernet
header. The MAC address of the equipment at both ends of the connection
is known to the BS and MS and need not be communicated.
I may have read this wrongly. If you are talking about the Ethernet
specific part of the packet CS in sentence 4 onwards (not the IP CS) ,
then the text is mostly ok, but unclear about what CS is being referred
to. Also the adding and removing of Ethernet headers is the role of the
Ethernet CS, not some additional behavior that is yet to be defined.
I suggest
“
2.2.2.3. IPv6 Transport
Note that in this scenario Ethernet CS may be more appropriate than
IPv6 CS to transport IPv6 packets, since the scenario needs to support
plain Ethernet connectivity end-to-end. However, the IPv6 CS can
also be supported. The MS and BS will consider the connections between
the peer IP CSs at the MS and BS to form a point to point link.
Every MS and the BS has the Ethernet type MAC
address. If the MS is using IP CS, then the BS needs to take care of
the Ethernet header. In the upstream direction, the BS will need to
generate an appropriate Ethernet header and prepend it to the IP
datagram. In the downstream direction, the BS will use the
destination address from the Ethernet header to identify the MS and
then it will strip the Ethernet header before relaying the IP
datagram over the 802.16 MAC connection. In the Ethernet CS case, an
Ethernet bridge may provide
implementation of authoritative address cache and Relay DAD.
Authoritative address cache is a mapping between the IPv6 address and
the MAC addresses of all attached MSs.
“
That’s all.
DJ