|
Gabriel,
The WiMAX Forum NWG reviewed section 4-3 and
Appendix C of draft-ietf-16ng-ipv4-over-802-dot-16-ipcs-04.txt and would like to
make a couple of remarks on the MTU issue:
- The I-D is not very
clear about the MTU issues appearing in an IPv4 over IEEE 802.16 transmission
system. The ambiguities mainly result out of the vague definition of the IEEE
802.16 link comprising both the radio part of the link as well as the part of
the link between BS and AR, when the functions are located in different
entities.
- Section 4.3 in
particular misses the discussion of the dependencies between MTU size and the
tunneling protocol deployed between BS and AR.
- Section 4.3 also
misses, that there is no packet loss when the MTU size limitation is caused by
the encapsulation overhead on the link between BS and AR. E.g. when GRE is used
for the tunnel between BS and AR, the transport IP layer can fragment the GRE
packets to fit the transport MTU on the link between BS and AR.
Reassembly in the tunnel endpoint at the AR will re-establish the original
user IP packet.
- Please note that
the reason of the WiMAX NWG to limit the MTU going over IPv4-CS to
1400 Bytes was to avoid fragmentation on the link between BS and ASN-GW as well
as on the link between ASN-GW and CSN (MIP tunnel). Fragmentation and
re-assembly require considerable processing power in the network
elements.
- Appendix C makes
statements which would require more detailed review of the I-D by WiMAX NWG. In
particular 'The addressing and operation of IPv4-CS described in this document
are applicable to the WiMAX networks as well' has not been verified
yet.
Furthermore 'Thus, WiMAX MS nodes should
use this default (1400) MTU value per the current specification [WMF].
However, due to reasons specified in section 4.3 above, it is strongly
recommended that future WiMAX MS nodes support a default MTU of 1500 bytes, and that
they implement MTU negotiation capabilities as mentioned in this document.'
makes recommendations to WiMAX without understanding the real reasons for
the limitation of the MTU size in Mobile
WiMAX.
We would recommend to 16ng to revise
the sections on MTU size to better explain the underlying issues leading to
restrictions in the MTU size. In particular the influence of tunneling inside
the network should be carefully discussed.
In addition we would kindly ask to either
remove whole Appendix C on the WiMAX MTU size or revise the text explaining the
real issues in the WiMAX architecture. In particular the statements on the
applicability of the I-D on the WiMAX architecture and the recommendation on
future modifications in the WiMAX architecture seem not to be very appropriate
to us.
Bye
Max
Vize Chair
NWG
From: ext Gabriel Montenegro [mailto:Gabriel.Montenegro at microsoft.com] Sent: Monday, November 03, 2008 11:59 AM To: nwg-chair at list.wimaxforum.org Cc: 'Daniel Soohong Park' Subject: [nwg-chair] NWG feedback on 16ng's IPv4 CS draft Prakash, Max and Yong Chang, |