MEXT WG @ IETF-79 Meeting Minutes
Meeting minutes based on notes taken by Jouni Korhonen
and Telemaco Melia.
MONDAY, November 8, 2010
0900-1130 Morning Session I
Valley Ballroom C INT mext
Mobility EXTensions for IPv6 WG
- Administrativia & Status Update, Chairs
Chairs: 3775bis is done, flow mobility is done. WG has
completed rechartering with new work items.
Two voluntary reviewers for MIPv6 alternative security
from Jouni Korhonen: Suresh Krishnan & Sri Gundavelli.
One voluntary reviewer for Behcet's draft on DSMIPv6
HA IPv4 address provisioning with DHCPv4: Kent Leung.
- Overlapping IPv4 Address Assignment Support for Dual-stack Mobile
IPv6, Sri Gundavelli
http://tools.ietf.org/html/draft-gundavelli-mext-dsmip-ipv4-overlap
Sri Gundavelli: Not explicitly supported in RFC5555,
but implicitly required. A kind of implementation issue. Needed by e.g. 23.975
IPv6 transition scenarios.
Raj Patil:This is purely internal to the
implementation, not protocol impact, why do we need to specify.
Behcet Sarikaya: no need for this. It is by design a
working scenario already.
Raj Patil: what needs to be extended in DSMIP to
support this?
Sri Gundavelli: have scenarios described and describe
what is needed in Binding Cache Entries and what options need to be there.There
is assumption that a HA gives unique addresses per MN
Raj Patil: agree on that.
Unknown: is this going to be purely informational? it
does not need to any protocol impact?
Sri Gundavelli: used today already.
Hidetoshi Yokota: It is deployed today in IPv4.
Three voluntary reviewers: Jouni Korhonen, Raj Patil, Christian
Gotare
- Security On Demand for Mobile IPv6 and Dual-stack Mobile IPv6,
Basavaraj Patil
http://tools.ietf.org/html/draft-bajko-mext-sod
Raj Patil: securing user data is optional in 3775, the
choice when use security is done by MN. Makes the selection of user data
protection negotiable.
Kent Leung: IKE is used for this. Why we need this for
MIP level?
Julien Laganier: IKE is negotiating the security
associations, not the security policy. This extension is about policy and how
to update the SPD.
Kent Leung: wants a usage scenario to be detailed.
Stephano Faccin: who decides in the MN? Also in 3GPP
there is no notion of trusted/untrusted networks that are based on roaming
contracts.
Suresh Krishnan: draft should be more clear why IPsec
mechanisms are not enough.
Two voluntary reviewers: Jouni Korhonen, Stefano Facin
- Authorizing Mobile IPv6 Binding Update with Cryptographically
Generated Addresses, Julien Laganier
http://tools.ietf.org/html/draft-laganier-mext-cga
Julien Laganier: Generates a HoA that is a CGA.Signs
BU with the private key. HA can have repository of authz MN pub keys, or only
authorize MN that are on-link at some point. No dependency on IPsec, no impact
on IPsec. Allows fully decentralized HA operation (e.g. for distributed
mobility management).
Four voluntary reviewers: Alper Yegin, Jean-Michel
Combes, Sri Gundavelli, Jouni Korhonen
- 3GPP TFT Reference for Flow Binding, Mohana Jeyatharan
http://tools.ietf.org/html/draft-jeyatharan-mext-flow-tftemp-reference
Julien Laganier: First, this is only useful in 3GPP so
there’s no reason to do it in IETF. 3GPP has Vendor Specific option and they
can do the work if they wishes. Second, in 3GPP TFTs are only used for QoS, so
you’d still need the MEXT specified traffic selector for traffic not subject to
TFTs. Thus is not worth the trouble to save 1 byte in corner cases.
No real support.
- NAT64 for Dual Stack Mobile IPv6, Behcet Sarikaya
http://tools.ietf.org/id/draft-sarikaya-behave-mext-nat64-dsmip
Various: Problem description in the draft is unclear.
- Extensions needed for MIPv6 to resolve 4G wireless gaps, Charles
Perkins
Charlie Perkins: 3GPP just did not select Mobile IP..
maybe 4G networking will. One proposal is allow GTP as a tunneling protocol,
another is to allow for Foreign Agent in MIPv6. Another is to specify different
tunnel endpoint for Home Agent, etc.
Kent Leung: how to deal with roaming?
Charlie Perkins: It’s a different topic.
Unknown: 3GPP arch already supports SGW/PGW
collocation.
THURSDAY, November 11, 2010
1740-1940 Afternoon Session III
Valley Ballroom C INT mext Mobility
EXTensions for IPv6 WG
- Agenda Bashing, Marcelo Bagnulo and Julien Laganier
Hui Deng: Are we defining the PS or the solution? DMM
is already in the charter? I have issues with the point in the agenda about the
solution
Marcelo Bagnulo: we want to discuss how current
protocols can solve some of the issues presented in the problem statement.
Hui Deng: I do not understand the agenda. It’s
solution space.
Marcelo Bagnulo: shall we move to the technical
discussion or continue to chat about the agenda? We are not doing solutions.
Rajeev Koodli: we are discussing about the problem
statement before we go to the solution, right?
Marcelo Bagnulo: Yes
- Problem statement for distributed and dynamic mobility management,
Dapeng Liu
http://tools.ietf.org/html/draft-chan-distributed-mobility-ps
Rajeev Koodli: what do you mean by mobility? I fail to
see to the problem statement.
Julien Laganier: let's go through the presentation and
keep the questions for the end
Parviz Yegani: what is the L-GW in this picture? Is
this the L-GW defined in the EPC?
Dapeng Liu: this is the local GW that we want to deploy;
this is the function that we want to distribute
Parviz Yegani: are you trying to describe a traffic
offload protocol?
Marcelo Bagnulo: let me provide some guidance at this
point, what we are trying to do is present a problem. It's not clear at this
stage if with the current protocols that we have could solve the problem
presented here
Telemaco Melia: Dapeng's point is that there are
already some deployment scenarios considering local anchoring
Kent Leung: I do not see the problem statement very
well
Marcelo Bagnulo: this is the discussion that I want to
have later
Raj Patil: when you have mobility all traffic is routed
through the home, this is not necessarily the case as in route optimization,
what is the mobility model that you have in mind?
Marcelo: please let’s have the questions at the end, let
me do my job, let me go through the presentation, else the we risk to not stick
to the agenda
- Use case scenarios for Distributed Mobility Management, Hidetoshi
Yokota
http://tools.ietf.org/html/draft-yokota-dmm-scenario
- Global HAHA solution, Ryuji Wakikawa
- Distributed Mobility Management with Proxy Mobile IPv6, Julien
Laganier
- End of the presentations and Discussion
Microphone to the chairs and opening the floor for
questions
Marcelo Bagnulo: what do you think about these
problems? Do you think they need to be addressed?
Behcet Sarikaya: clarification question to Julien, in
3GPP where would be your HA?
Julien Laganier: you could map this to 3GPP if you
want but let's not discuss 3GPP specific
Sri Gundavelli: clarifying question, what is the
impact of this proposal on legal intercept, charging and accounting. Can you
clarify?
Marcelo Bagnulo: what is the goal of your question?
Sri Gundavelli: we need to understand the impact on
other services.
Hidetoshi Yokota: are you talking about the control
function?
Sri Gundavelli: How you are going to perform billing
for instance?
Hidetoshi Yokota: if the mobile network allows for
distribution then we can distribute the function and other services will be
adapted to meet the requirement
Hui Deng: if you want to do billing/lawful
interception, then you don’t need to do breakout, you could continue to go
through mobile core network, this is what 3GPP SA1 specified already
Christian Gotare: I am missing some points, we need to
account for the number of handoffs, the residency of the MN in a cell, etc…
Marcelo: are you asking data about MN distribution?
Christian: we need to discuss how people move, how
this has an impact on the distribution?
Parviz Yegani: we need to be positive about what is
happening here, the anchor points have to be located close where the MN are
located, this new architecture has merits but we need to check if there is
merit to develop solutions in the IETF, if there is a need for low level
latency we need to understand this
Serge ?: this is problem is very interesting, service
providers are looking at this, and this is justified by the LIPA SIPTO work
Jari Arkko: this has an impact on the legal interception;
it is an implication that needs to be taken into account
Sri Gundavelli: can we simply assume that we are going
to introduce another level of complexity and cost?
Jari Arkko: need to identify the right tradeoffs about
costs
Suresh Krishnan: the closer you put the anchor to the
MN the more handoff happens and as such signalling increase, we need to find
the right balance
Parviz: the services are not equal, some services might
not require seamless mobility, they are in the bottom part of the value chain
Sri Gundavelli: what are the missing functions to
answer to the problem statement?
Hidetoshi Yokota: we do not talk about protocols in
the presentation.
Sri gundavelli: to support this architecture what are
the extensions required?
Jari Arkko: we do not have an answer. We need to look
at use cases and then at the existing protocols, it maybe sufficient it may not
Sri Gundavelli: ok you are looking at understanding
Rajeev Koodli: you are mostly looking at deployment
choices, control data plane separation, we did look at local routing, what
might seem reasonable is taxonomy of what the choices are, this is system design
not protocol design, the industry should choose the right tool box
Suresh Krishnan: first we need to understand the problem,
then we look at the protocol semantics, we need to watch out for the changes, but
first of all agree on the problems
Raj Patil: good to see a sort of scenarios and pain
points that have been presented, it is very useful, the reason why they bring
this up is a matter of costs, the scenarios are useful, the point of DMM is not
yet clear though, it is also a fact that mobile operators are deploying local
anchors because there is a benefit, operational costs is another key point to
consider, rather than talking about PMIP or MIP I would prefer to look at the
problem as a whole, way to early about solutions
Charlie Perkins: agree on what Raj said, it is critical
to understand what can be done today with existing solutions: once we do this,
we will know what part/extensions belong to the network and which part to the
mobile, after this we will be able to identify which part of the solutions will
solve what
Hidetoshi Yokota: if CN talks for the first time, it
has to find to which mobility anchor to talk to
Christian Gotare: comment about physical realization;
how many people; how many stations, etc
Jari Arkko: answer about what Suresh said, I think he
is right, I think that new code might be needed on the mobile node, monitoring
to decide which address is used
Ahmad Muhanna: probably what we are missing is an
algorithm to dynamically assign an anchor
Telemaco Melia: it’s true services are not equal, some
need to be seamless; some others not: as a company are looking at an interface
between applications and the network; that’s why we introduced the dynamic feature
Ruji Wakikawa: I need more pictures about the network
environment, I need to know about user mobility, we want to optimize path but
we need to know how fast the user is moving since we could end up with a longer
path
Rajeev Koodli: distribution of mobility anchors, let's
folk decide how do they want to deploy the network, I do not see on what
protocol we need to work, what we need to resolve other than understanding
Hong Liu: 3GPP is specifying SIPTO, here is just copy
what they are specified, and used in IP mobility
Jari Arkko: can we get the list of problems that 3GPP
came up with? It is important to understand the contexts as well.
Hui Deng: 3GPP SA1 has requirement for
SIPTO, which could be used here for reference.
Gaetan Feige: we should look at the type of service
and users other than simply distribution of the mobility management, traffic characterization
is important
Hui Deng: yes, the traffic pattern is different from
mobile Macro station before. HNB or enterprise HNB could have different traffic
pattern, and for SIPTO case, there are also different traffic pattern.
End of the discussion – microphone again to the chairs
The following questions have been asked to the room:
- are people interested in this work and to define the
problem? (nobody did oppose)
- are people interested in looking at if current
technologies are sufficient to solve the problem? (nobody did oppose)
Jari Arkko: given the positive feedback of the
audience, first we identify the problem then understand if current solutions
are sufficient. Documents will be produced to reflect this decision.