RE: [Seamoby] Minutes for Meeting at IETF 53

["Gary Kenward"<[e-mail]>]

My observation on this MN centric approach is that, to me, there
are logical flaws. Let me present these flaws in two cases; I will
refer to the "old and candidate" networks, meaning the network that
was supporting the MN, and a network that is a possible candidate
network for handover of the MN's traffic. Each has an AR, AP, etc.

  1. The old and candidate networks have an interworking capability:
     in this case, the networks have a more complete, and current
     view of the services available before and after handover. The
     proposal is to use CAR to ship information describing this view
     to the MN. At best, this implies using wireless bandwidth, and MN
     CPU cycles and battery power to make a decision that can be made
     by two networks that have relatively unlimited link bandwidth, 
     CPU cycles, and power.

     Note that if the two networks are under the same administration,
     it would seem reasonable to assume that the networks are deployed
     with some sort of interworking capability.

  2. The old and candidate networks have no interworking capability,
     e.g. they know nothing of each other. In this opposite extreme,
     the primary model is essentially one of an IP service running 
     over two wireless access networks. Since, by definition, the MN
     is the main common element between the two networks, it would 
     appear to make sense to have it make the handover decision.

     There are then three alternatives to how the MN can access the 
     information from the candidate network: the candidate network
     broadcasts capability information "regularly" over the wireless
     channel; the candidate network responds to all requests for
     capability information; and, the candidate network responds
     to authorized requests from a specific MN (implying that the
     MN is authenticated). 

     Now, at this point, I will assume that each network has an
     operator that wishes to recover his/her investment in spectrum,
     equipment, backhaul links, operating costs, etc. In a commercial
     operation, it is reasonable to assume competition: what operator
     would be willing to promiscuously broadcast capability information
     about their network that could be used by a competitor to lure
     away customers? Wireline network operators are leery of exposing
     any information that might identify their network topology or
     how it operates, and so why would a wireless operator be different?

     So, this implies that the MN must establish a link with the candidate
     network, be authenticated and authorized to receive capability
     information. A lot has to happen before IP packets can be exchanged
     (CAR is an L3 solution), and, in order for capability discovery
     to be useful, the MN must be in communication with at least two
     candidate networks. If the MN is assumed to have single channel, 
     multi-mode NICs, this discovery process has to occur sequentially 
     - the equivalent of channel scanning used in FDM wireless systems 
     - which is very difficult to do and provide "seamless" handover,
     particularly when L2 authentication and L3 capability discovery
     have to be completed. 

     This leads to the conclusion that the MN has to have multi-channel, 
     multi-mode NICs, which might make the NIC vendors happy, and
     certainly would make the battery vendors very happy.

None of these arguments apply if we are talking about best effort service
over an "open" access network. To me, the idea of must be open and
non-commercial
access networks outside of select government/academic networks is
unrealistic.
It is arguable that that access to the Internet today is predominately 
commercial.

The bottom line is that there are significantly greater resources within
the network to perform these functions, and the communications between
networks are relatively (note, I said "relatively") easier to make secure.

Why does this matter? If Seamoby develops both solutions, then won't the
marketplace determine what makes sense or not? Perhaps, but this approach
makes sense only if there is no cost to defining a one-size-fits-all
solution.
I would prefer that the effort be focussed on inter-AR capability discovery,
including internetwork capability discovery. 

Just more of my delusional thoughts,
Gary

> -----Original Message-----
> From: Karim El-Malki (ERA) [mailto:[e-mail]]
> Sent: April 18, 2002 10:09
> To: [e-mail]
> Subject: RE: [Seamoby] Minutes for Meeting at IETF 53
> 
> 
> Since a couple of emails disagreed with me about this point,
> I'm sending one email reply to clarify. As in the minutes,
> the concept was that the only entity which can find out
> all the possible CARs is the MN (that's what I mean by 
> mobile-centric).
> The network could give hints, but it is the MN that has the complete
> view and decides. That's what I understood from the meeting and there
> seemed to be consensus on this. Hope we agree so far. The network
> could assist in CAR discovery (e.g. it's L2 can help in anticipating
> the movement). That's also somewhere in the minutes. In fact what I
> said below was that I couldn't understand why there was focus on a
> network-centric solution (i.e. where the MN is NOT involved). I wasn't
> ruling out network involvement, but I couldn't understand why there
> was work on ARs only doing CAR discovery. Hope my point is more clear
> now. Maybe I misinterpreted some emails and people are in agreement
> with this. I think Dirk's email at least was in line with what I
> have above (i.e. we need to focus on the MN's role and allow the
> network to assist where needed).
> Rgds
> /Karim
> 
>  > -----Original Message-----
>  > From: Govind Krishnamurthi [mailto:[e-mail]]
>  > Sent: den 17 april 2002 21:22
>  > To: [e-mail]; [e-mail];
>  > [e-mail]; [e-mail]
>  > Subject: RE: [Seamoby] Minutes for Meeting at IETF 53
>  > 
>  > 
>  > 
>  > 
>  > 
>  > Hello Karim,
>  > First, I don't think that Steve Deering said anything 
> about excluding
>  > the network from CAR discovery. All he said (AFAIK) was 
> that the MN's
>  > preferences must be satisfied in the decision of selecting the TAR.
>  > Now that doesn't imply MN based CAR discovery: that the MN 
>  > should receive 
>  > all the capabilities and GAAR IDs and then make the decision.
>  > 
>  > 
>  > Regards,
>  > Govind.
>  > 
>  > >
>  > >The reason I sent my previous email was that I got the feeling
>  > >from this thread that there is focus on providing a solution for
>  > >the network-centric case (i.e. the network handles CAR discovery
>  > >and the MN isn't involved). I wanted to point out that this
>  > >solution would not be applicable to the mobile-centric case and
>  > >clarify that this case is supported by MIP (v4 and v6) fast
>  > >handoffs.
>  > >
>  > >From the IETF meeting there was a majority who wanted to go for
>  > >the mobile-centric case (i.e. where the mobile is involved and
>  > >decides). Following Steve's presentation I think that became
>  > >quite clear. So, why are there discussions about doing a solution
>  > >which is meant for the network-centric case only? This solution
>  > >wouldn't apply to the mobile-centric case. That's what I was meant
>  > >to ask in my last email.
>  > >
>  > >/Karim
>  > >
>  > >  > Karim,
>  > >  >
>  > >  > But it is applicable to network initiated handoff. And 
>  > there are two
>  > >  > algorithms, one in which the access router sends a PrxyRtAdv
>  > >  > to the MN
>  > >  > and one in which the MN is switched (for FMIPv4).
>  > >  >
>  > >  >             jak
>  > >  >
>  > >  > ----- Original Message -----
>  > >  > From: "Karim El-Malki (ERA)" <[e-mail]>
>  > >  > To: "'James Kempf'" <[e-mail]>; 
> "Hemant Chaskar"
>  > >  > <[e-mail]>; <[e-mail]>
>  > >  > Sent: Tuesday, April 16, 2002 11:55 AM
>  > >  > Subject: RE: [Seamoby] Minutes for Meeting at IETF 53
>  > >  >
>  > >  >
>  > >  > > > > >[The Issue]
>  > >  > >  > > >It seems that the minutes indicate that people have
>  > >  > >  > somehow come to a
>  > >  > >  > > >conclusion that there is no need for access routers to
>  > >  > >  > divulge that
>  > >  > >  > they
>  > >  > >  > > >are
>  > >  > >  > > >geographically adjancent to themselves via the IP
>  > >  > >  > infrastructure. The
>  > >  > >  > > >minutes also seem to indicate that the mobile 
>  > alone should be
>  > >  > the
>  > >  > >  > only way
>  > >  > >  > > >to pass addresses of CARs to source ARs.
>  > >  > >  > > >
>  > >  > >  > > >[Technical Questions]
>  > >  > >  > > >O.K. If this is true then how will a mobile pass this
>  > >  > >  > information to
>  > >  > >  > their
>  > >  > >  > > >source AR if the mobile only has one NIC and 
> that NIC is
>  > >  > >  > only capable
>  > >  > >  > of
>  > >  > >  > > >listening to one media at once?
>  > >  > >  > >
>  > >  > >  > > [HC] I agree that this is a genuine technical 
>  > problem. What I
>  > >  > would
>  > >  > >  > really
>  > >  > >  > > like to understand is whether the above is a 
>  > relevant case for
>  > >  > CAR
>  > >  > >  > discovery
>  > >  > >  > > or we simply neglect it and focus only on two
>  > >  > physical interfaces
>  > >  > >  > case. I
>  > >  > >  > > raised this question before, but we have not had much
>  > >  > discussion
>  > >  > on
>  > >  > >  > it. In
>  > >  > >  > > any case, address translation part of CARD will 
>  > be required in
>  > >  > this
>  > >  > >  > case for
>  > >  > >  > > fast handoff support.
>  > >  > >  > >
>  > >  > >  >
>  > >  > >  > In a single interface handoff situation, Layer 2 
> typically
>  > >  > >  > delivers the
>  > >  > >  > AP or AR L2 identifier to which the MN will be 
>  > handed over. This
>  > >  > >  > information is required (by the MIP fast handover 
>  > algorithms) at
>  > >  > the
>  > >  > >  > MN's AR. So the issue is fairly simple: the AR must be
>  > >  > able to do
>  > >  > >  > reverse address translation in order that it can 
>  > contact the
>  > >  > >  > other AR.
>  > >  > >
>  > >  > > This is not really applicable to Mobile-Initiated MIP 
>  > Fast Handoff.
>  > >  > > In Mobile-initiated we consider the useful case where 
>  > the MN can
>  > >  > recover
>  > >  > > the CAR IP address/es from the L2 trigger. The MN then
>  > >  > sends a Proxy
>  > >  > > Router Solicitation (RtSolPr) to its curent AR containing
>  > >  > one or more
>  > >  > CAR
>  > >  > > IP addresses. This means that the source AR already 
>  > gets the CAR
>  > >  > addresses.
>  > >  > > So we can do without the AR doing translation or discovering
>  > >  > geographical
>  > >  > > adjacency. The MN can pass these CAR address/es to the
>  > >  > source AR for
>  > >  > both
>  > >  > > single and multiple-interface MNs. So I think that the
>  > >  > conclusion from
>  > >  > the
>  > >  > > meeting is compatible with MIP Fast Handoffs.
>  > >  > >
>  > >  > > /Karim
>  > >  > >
>  > >  >
>  > >
>  > >_______________________________________________
>  > >Seamoby mailing list
>  > >[e-mail]
>  > >https://www1.ietf.org/mailman/listinfo/seamoby
>  > 
>  > 
>  > _________________________________________________________________
>  > Chat with friends online, try MSN Messenger: 
http://messenger.msn.com
 > 

_______________________________________________
Seamoby mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/seamoby