Re: [netext] Comments on :draft-wu-netext-local-ro-04.txt

["Mohana Jeyatharan" <[e-mail]>]

Hi Qin,

Thanks for reply.  Please see some replies in-line.

BR,
Mohana

> -----Original Message-----
> From: Qin Wu [mailto:[e-mail]]
> Sent: Friday, November 20, 2009 10:19 AM
> To: Mohana Jeyatharan
> Cc: [e-mail]
> Subject: Re: Comments on :draft-wu-netext-local-ro-04.txt
> 
> Hi, Mohana:
> Sorry for my late reply. Also thank you for your review and comments.
> please see my reply inline.
> 
> Regards!
> -Qin
> ----- Original Message -----
> From: "Mohana Jeyatharan" <[e-mail]>
> To: "Qin Wu" <[e-mail]>
> Cc: <[e-mail]>
> Sent: Monday, November 16, 2009 9:59 AM
> Subject: Comments on :draft-wu-netext-local-ro-04.txt
> 
> 
> 
> Hi Qin,
> 
> Read your ID.  I have few comments.
> 
> Please see below.
> 
> BR,
> Mohana
> 
> In abstract,
> "In case the correspondent node is
>     connected to another local mobility anchor, the local mobility
>     anchors connected by the correspondent node needs to be discovered
>     firstly so that it can notify its mobile access gateways attached
by
>     the correspondent node to the mobile access gateway attached by
the
>     mobile node afterwards."
> 
> <<<[Mohana] In abstract it is not very clear which entitiy notifies
CN's
> MAG to MN's MAG. Perhaps a rewording may be useful>>>
> 
> [Qin]: Yes, what we are really saying here is CN's LMA will notify
CN's
> MAG to MN's MAG when MN's MAG
> communicate with CN's LMA using normal PBU/PBA. Maybe the proper way
to
> say would be
> "
> so that the discovered local mobility anchor notity its mobile access
> gateways attached by the correspondent node......
> "
[Mohana] So, the description is tied to CN's LMA notifying MN's MAG to
CN's MAG. Ok, understand.  But some description on MN's LMA notifying
CN's MAG to MN's MAG will be also useful in the inter-LMA scenario.
What I am not clear is why is MN's MAG connected to CN's LMA?  Probably
this is another scenario addressed in scetion 7(inter LMA case).  Anyway
I will leave it for now.  If MN's MAG is connected to CN's LMA, then
such passing (CN's LMA passing CN MAG address to MN's MAG)is propoer.
Otherwise Cn's LMA will need to pass CN's MAG address to MN's LMA.

> In abstract, Mobile access gateways create and refresh bindings
>     using proxy binding update and acknowledgement messages.
> 
> <<<[Mohana] What is the reason for this.  Is this refering to PBU to
LMA
> or PBU to MAG?  Perhaps more clarity on this will help.>>>
> 
> [Qin]: Okay, actually PBU/PBA will be used between MAGs to create
binding
> when it is the first time for MAGs to setup localized routing path.
> during handover case, PBU/PBA will be used between MAGs to refresh
binding.
> The reason to create and refresh bindings between MAGs attached by
mobile
> node and correspondent node is quite similar to correspondent
registration
> in the Mobile IPv6 Routing optimization between mobile node and
> correspondent node described in the section 11.7.2 of RFC3775.
> i.e., allowing correspondent node to cache mobile node's proxy care-of
> address or allowing mobile node to cache correspondent node' proxy
care-of
> address,
[Mohana]I agree with your approach of using PBU/PBA for routing state
setup and tunnel setup between MAGs and think that is a good approach
and tied to RFC 5213 way.  From further reading of your draft I
understand this.  All I wanted to say is that in abstract, a mentioning
of PBU/PBA from MAG to MAG introduces more clarity.  That is all.
> 
> In conventions used section it is mentioned
> 
> Local routing: Traffic between MN and CN does not pass through LMA
>    and is locally routed in the same PMIPv6 domain.
> 
> <<<[Mohana] Perhaps a re-wording may be necessary when referring to
> PMIPv6 domain. This should be aligned with the PS draft terminology.
It
> should be same provider domain running PMIPv6 or some thing like
> that.>>>
> 
> [Qin]: Okay, will check the wording to be consistent with PS draft.
> 
> In conventions used section it is mentioned
> Local Routing Optimization Request (LROREQ): A message initiated by
>    the MAG or LMA requesting the MAG or LMA to establish local routing
>    optimization path between MN and at least one pair of CNs who
>    communicate with MN.
> <<<[Mohana] why is there reference to at least one pair of CN.  Are u
> saying at least a single CN? Not very clear>>>
> 
> [Qin]: Right, thank for your good catching and correction.  The reason
to
> say at least one CN is
> in some cases, maybe MN will communicate with more than one CN, in
this
> sense, LROREQ
> message may be sent from MN's LMA to each MAGs attached by serveral
CNs
> which is communicating with MN.
> or sent from MN's MAG to each LMA which is registered by more than one
CNs
> which is communicating with MN during inter-LMA handover case.
> 
[Mohana]Ok, sure I understand your intention proabably a minor
re-wording would do.

> In section 3 Scenario analysis for PMIPv6 local routing
> 
> <<<[Mohana]The text before the figure 1 is refering to PMIPv6 domain.
> Again I think this should be single administrative/provider domain
> running PMIpv6.>>>
> 
> [Qin]: Okay.
> 
> In figure 1,
> <<<[Mohana]the MAG1 is connectedto IPv4 only or IPv4/IPv6 transport to
> LMA.>>>
> 
> [Qin]: I am okay with this change in the figure 1.
> 
> In last para in section 3 which is
> In all the above scenarios, MN is allowed to roam within its PMIPv6
>    domain (i.e, MN's home domain in which MN's LMA is located) or move
>    from one PMIPv6 domains(i.e., MN's visited domains)to another. CN
>    should stay with MN within the same PMIPv6 domain, e.g., MN moves
to
>    the visited domain to which CN belongs. Another example is MN and
CN
>    move to the same visited domain to which MN's LMA or CN's LMA does
>    not belong.
> <<<[Mohana] Here PMIPv6 domain should be chnaged to administrative
> domain. Probably re-writing with respect to MN in home or visited
domain
> and CN inhome or visited domain may be useful.  Currently the text is
> not very clear>>>
> 
> [Qin]: Agree there is room for improvement here, this paragraph will
be
> reworded in accordance with the
> terminologies used in the PS draft
[Mohana] Ok, thanks.

> In section 4,
> <<<[Mohana] the first bullet should be changed to administrative
domain
> instaed of PMIPv6 domain.  The second bullet says MN and CN attached
to
> same LMA.  But in section 3 it is mentioned inter LMA scenario as
well.
> So, I am a bit confused as to whether the solution is not handling
inter
> LMA scenario.>>>
> 
> [Qin]: Yes, inter LMA scenario will be handled in one separate section
of
> our Solution I-D. Our solution I-D mainly focuses on specifying the
> localize drouting protocol based on the first two scenarios described
in
> the section 3. the common aspect of the first two scenarios is it does
not
> require PMIP6 mobility entities discovery described in the PS draft.
As
> regarding to the inter-LMA local routing scenario, it depends on PMIP6
> mobility entities discovery, extra extended PMIP6 signaling is
required
> between MN's MAG and CN's LMA. Therefore inter-LMA local routing case
is
> separated from section 4 which is concentrating on the first two
scenarios
> and addressed in the independent section, i.e., section 7.


[Mohana]  Ok, understand. Do not know why I got confused.  Probably a
section outline may be helpful and I leave it to you to handle.
> 
> In section 4.1
>  <<<[Mohana] In this section there is reference to three different
types
> of flags such as intra-LMA local routing flag, intra-MAG local routing
> flag, Enable detect local routing flag.  A description of these in
> section 2 would be better for easy understanding.>>>
> 
> [Qin]: Okay. Thank for your suggestion.
> 
> In section 4.1, para 3, it is mentioned After successful local routing
> optimization process, if MN and CN
>    attach to the different MAGs, i.e., MAG1, MAG2 and intra-LMA
>    localrouting flag is set to value one, the MAG1 to which the MN
>    attaches may send PBU message to the MAG2 which the CN attaches to.
> <<<[Mohana]Here I think there needs to be a rephrasing of sentence.  I
> think you are referring to MN connected to MAG 1 and CN connected to
> MAG2 and intra LMA flag is set at the MAG and PBU/PBA exchange between
> th etwo MAGs.  The sentence is not clear.>>>
> 
> [Qin]: Okay, maybe we could say:
> "
> MAG1 to which the MN is connected may send PBU message to the MAG2 to
> which the CN
> is connected.
> "
[Mohana] probably somewords along the above lines may be useful.
> In section 4.1, in para 3 it is mentioned that MN's MAG is informed of
> CN MAG via LROResponse from LMA (assuming MAG initiated RO) and CN's
MAG
> is informed of Mn'sMAG via LRO response and both MN's Mag and CN's
MAGs
> are engaged in bi-ditectional PBU/PBA.
> 
> <<<[Mohana]Here I have some comments.
> Mn's MAG will do LROReq to LMA. LMA can pass CN's MAG address in new
> option.  Then MN's MAG can start PBU with CN's MAG.  But the question
I
> have is how does CN's MAG create the routing state for MN prefix.  How
> does the CN's MAG know that this prefix is owned by MN and MN's MAG is
> actually proxying that prefix?
> 
> 
> The same when CN's MAG does PBU to MN's MAG.  Probably, after getting
> the PBU from MN's MAG, CN's MAG can start PBU with CN's MAG.  But
again
> to create the the routing state for CN's prefix at MN's MAG the MN's
MAG
> should know the prefix CN owns is valid prefix. Otherwise, I don't see
> how a valid routing state can be created.
> 
> So, by reading this I feel a prefix ownership kind of information is
> needed in the PBU.>>>
> 
> [Qin]: My understanding is if CN's MAG trusts MN's MAG, then CN's MAG
must
> believe the prefix sent from MN's MAG is owned by MN. Because MN's MAG
can
> check whether prefix is ownd by MN and used to create a valid routing
> before it send the prefix to the CN's MAG. That is to say we can
assume
> there is a trust relationship between MN's MAG and CN's MAG. In oder
for
> this, two possible ways in my mind are,
> 1) we assume a prior agreement between MN's MAG and CN's MAG.
> 2) the second way is we use IPSEC to setup Security association
between
> MN's MAG and CN's MAG.
> By these two ways mentioned above, we can have trust relationship
between
> MAGs, But what if MN's MAG establish trust relationship with CN's MAG
and
> send the forged MN's prefix to CN's MAG, I think the prefix ownership
> validation may be useful in this scenario to help create valid routing
> state at CN's MAGs
> Therefore I would like to address this security threat in our solution
I-
> D.Thanks for your good points.


[Mohana] Thanks Qin.  Although I liked the Pbu way of establishing
tunnel between MAGs, this prefix validation seemed important to address
the security part. In RFC 5213, LMA is the prefix delegator so such
issue is not needed.  But here, the CN's MAG need such validation.
Thanks for addressing this part.
> 
> In the handover mechanims in section 4.1.1, it is mentioned that the
> nMAG1 will get context from old MAG (pMAG1).
> 
> <<<[Mohana] But the question I have is when nMAG1 does PBU to CN's
MAG,
> how can CN's MAG know that this prefix of MN is valid and nMAG1 is
> proxying the preifx.  Bcos during handoff, the LMA is not involved.
> nMAG1 has no issue in knowing that MN;s prefix is given by LMA because
> when it does PBU to LMA it will know that this is a valid prefix.  But
> that is not the case with CN's MAG.>>>
> 
> [Qin]: Good question, when MN hands over from pMAG1 to nMAG1 and keeps
on
> communication with CN, there is no trust relationship beforehand
between
> nMAG1 and MAG2 to which CN is connected.  That is true. But I think
MN's
> prefix will keep unchanged during handover. The only change to MN is
MN's
> Proxy CoA will . Is it necessary/mandatory to ask CN's MAG to validate
> MN's prefix ownership each time MN hands over? In my understanding,
prefix
> ownership validation can be understood as routing reachability
test(RRT)
> procedure defined in the RFC3775. But I am not sure RRT procedure will
be
> done each time MN hands over according to RFC3775.

[Mohana] For the handover case, just as in new establishment, the CN
needs to know that the nMAG is proxying the prefix.  So the prefix
validation method as for new connection can be used here. The RFC 3775
way of prefix validation is one way but there is more signaling.
Probably some assurance from LMA will be good regarding prefix validity.
I think this is critical bcos in handover case, LMA is not involved. So
to assure fast handoff one would not want the LMA to be involved.
Furthermore, using your method a distributed handoff management can be
done without the involvement of pMAG of MN notifying anything to CN'MAG.
What I am saying is nMAG of MN can handle the PBU with CN's MAG without
old MAG of MN being involved and this will expedite handover state
convergence.  The only thing that is needed in MN's nMAG to tell CN I
own/proxy this prefix and prefix is valid.

> Also in section 4.1 in second paragraph it is mentioned that MAG can
> send LRO req to LMA when Mn and CN are not attached to same MAG.
> <<<[Mohana] In this case, how does MAG know the CN address.  What
value
> will be attached in the Mn-CN pair mobility option.  Is there some L2
> message sent from Mn to MAG and all CN addresses are given in these.
> Probably such mentioning will be useful>>>
> 
> [Qin]: How MAG know the CN address actually is beyond scope of this
> solution I-D. The possible way is MAG may look at data packet'
destination
> address to know CN address.For the case where one MN communicates with
one
> CN, MN-CN pair mobility option may include one MN's info and one CN's
info.
> MN's. For the case where one MN communciates with several CNs, MN-CN
pair
> mobility option may include one MN's info and several CNs' info.

[Mohana] Ok, thanks. I was just analyzing the benefits of MAG initiated
RO as opposed to LMA initiated RO and wanted to highlight some benefits
when MAG initiates RO.  I guess how MAG gets the information need not be
covered in the draft and it can be out of bound signaling.

> I have more comments.  But rather than sending all at once I thought I
> will stop here for now.

> Thanks.
> 
> > -----Original Message-----
> > From: [e-mail]
> [mailto:[e-mail]]
> > On Behalf Of [e-mail]
> > Sent: Monday, October 26, 2009 4:30 PM
> > To: [e-mail]
> > Subject: I-D Action:draft-wu-netext-local-ro-04.txt
> >
> > A New Internet-Draft is available from the on-line Internet-Drafts
> > directories.
> >
> > Title           : Proxy MIP extension for local routing
> optimization
> > Author(s)       : W. Wu, B. Sarikaya
> > Filename        : draft-wu-netext-local-ro-04.txt
> > Pages           : 32
> > Date            : 2009-10-26
> >
> > This document extends local routing in proxy Mobile IPv6 and defines
> >  a localized routing optimization protocol within one PMIPv6 domain.
> >  The protocol supports IPv4 transport network operation, IPv4 home
> >  address mobility and handover. The Local mobility anchor/mobile
> >  access gateway initiates local routing for the mobile and
> >  correspondent node by sending messages to each mobile access
> >  gateway/local mobility anchor. In case the correspondent node is
> >  connected to another local mobility anchor, the local mobility
> >  anchors connected by the correspondent node needs to be discovered
> >  firstly so that it can notify its mobile access gateways attached
by
> >  the correspondent node to the mobile access gateway attached by the
> >  mobile node afterwards. Mobile access gateways create and refresh
> > bindings
> >  using proxy binding update and acknowledgement messages.
> >
> > A URL for this Internet-Draft is:
> > http://www.ietf.org/internet-drafts/draft-wu-netext-local-ro-04.txt
> >
> > Internet-Drafts are also available by anonymous FTP at:
> > ftp://ftp.ietf.org/internet-drafts/
> >
> > Below is the data which will enable a MIME compliant mail reader
> > implementation to automatically retrieve the ASCII version of the
> > Internet-Draft.