Re: MN behaviors to uplink data packets during a handoff

Hi Jay,

You are quite right. So after several times try and fail, link layer drops
the packet.

Another question is in OPNET MIPv4 model, what do "the last known agent" and
"the last known good interface"  mean? If the MN gets an agent advertisment
from a new FA, while haven't got any adversitments from its old FA recently,
then the new FA immediately becomes the last known agent? If so, at this
time the MN turns to send packets to the new FA without waiting for the
registration reply, and whether these packets from the unauthorized MN would
be served depends on whether the new FA would provide service to them?

Many thanks,
Jee


----- Original Message -----
From: "Jay J Lee" <[email protected]>
To: "Jee J.Z." <[email protected]>; <[email protected]>
Sent: Monday, July 28, 2003 2:02 AM
Subject: Re: [mobile-ip] MN behaviors to uplink data packets during a
handoff


> Hi,
>
> I am not an expert on WLAN behavior but I do not think it should hold
> packets when it looses connectivity with the AP.  The main reason for that
> is that this type of condition is best handled (at least currently) by the
> upper layer (most likely transport layer).
>
> Let's say you have a TCP session going on and you lost connectivity with
> your AP for 10 seconds then reconnects.  If the TCP session was terminated
> then the MAC layer is left with stale data that nobody can process if it
> were queueing packets (for other than the purpose of limited retries).
With
> UDP, you most likely want to receive the most current packets transmitted
> rather than the ones transmitted 10 seconds ago (real time video anyone?).
>
> Again, I am strictly talking about real life devices and their
> implementations.  I have seen similar proposals (smarter link layers) in
> several research papers. They were not necessarily related to Mobile IP
but
> on improving wireless network performance in general.
>
> Good luck, - Jay
>
>
> At 05:03 PM 7/24/2003 +0100, Jee J.Z. wrote:
> >Hi Jay and all,
> >
> >Thank you very much for your reply.
> >
> >According to your comments, can I think this way that the disruption to
> >uplink stream caused by a handoff is largely depending on the link layer
> >technology and the Mobility Agent (or the administrative domain)
behaviors?
> >That is,
> >
> >For example, as far as I know, Mobile Nodes in WLAN make sure there is an
> >Access Point reachable before everytime they send out a frame. In this
case,
> >even if the MN loses contact with any AP for some time, IP packets would
not
> >be lost, instead, they queue up at the link layer and wait for the link
data
> >frame to be sent out. (?)
> >
> >When the MN is in registration process with a new FA, it sends out data
> >packets to the new FA as long as it knows the FA's MAC address. Whether
to
> >offer service to the unauthorized MN (at least before the successful
> >registration reply comes back) depends on the policy the new FA (or the
> >admistrative domain it belongs to) deploys. If it does not, uplink data
> >packets are discarded by it during the registration process. If it does
> >offer service, there could be no uplink data packets lost in WLAN. (?)
> >
> >However, it is not WLAN, things might be different. (?)
> >
> >Regards,
> >Jee
> >
> >
> >----- Original Message -----
> >From: "Jay J Lee" <[email protected]>
> >To: "Ji Zhang" <[email protected]>; <[email protected]>
> >Sent: Thursday, July 24, 2003 3:59 AM
> >Subject: Re: [mobile-ip] [Mobile IP] MN behaviors to uplink data packets
> >during a handoff
> >
> >
> > > Hi,
> > >
> > > I believe these are all implementation time decisions.
> > >
> > > For #1, if you are talking about MIPv4, what is happening within the
IP
> > > layer should not affect the link layer.  If it still has upper layer
> > > packets to send out it will continue to do so.  The link/MAC layer
does
> >not
> > > and should not know what happens at the other layers other than the
> >packets
> > > it receives from them.  Although with MIPv6 there are proposals for
> > > inter-layer communications.
> > >
> > > For #2, the OPNET MIPv4 simulation model, for example, keeps sending
the
> > > data packets to the last known agent (HA or FA) on the last known good
> > > interface whether it is lost or in the process of registration.  This
> > > decision was made because once you start queueing packets and managing
> > > them, the implementation becomes expensive considerably in the
simulation
> > > domain as well as for the OSes in the real devices.
> > >
> > > Hope this helps, - Jay
> > >
> > >
> > > At 09:37 PM 7/23/2003 +0100, Ji Zhang wrote:
> > > >Hi all,
> > > >
> > > >Could anybody offer me some instructions on the behaviors of a Mobile
> >Node
> > > >to
> > > >uplink (MN-->FA) data packets during a handoff?
> > > >
> > > >For example,
> > > >1. when the MN loses link contact with the old FA (before getting an
> >agent
> > > >advertisement), are all uplink data packets stored up at the link
layer?
> > > >2. when the MN is aware of agent advertisements are missing, at the
same
> > > >time it sends an agent solicitation, should it stop sending data
packets
> > > >until
> > > >it gets an agent advertisement (or even until the registration is
> > > >successful)?
> > > >
> > > >Many thanks,
> > > >Jee
> > >
> > >
>
>