Hello,

On Wed, 2006-09-06 at 08:25 -0400, Thomas Narten wrote:
> Pars Mutaf <[EMAIL PROTECTED]> writes:
> 
> > Thanks. But I still believe that a host should be able to test 
> > if it is reachable in dormant mode (reachable, or reachable within
> > reasonable delay). This is good for dormant mode security.
> 
> Why? This is a problem statement you are asserting. Rather than
> discuss possible solutions, I think it is important to first establish
> whether this is an actual problem people think we need to solve.
> 

OK.

When the host is dormant, it is reachable through a different 
set of links. Because, the dormant mode state must be checked 
somewhere in the system, etc. There is also the paging channel.
(different, and time-slotted). Paging has a complex architecture.
It is more complicated than active mode support. An IP-layer 
solution can check all of these and detect failure: From AR
to the host, everything between the two will be tested.

IMHO, we should not assume that the cellular system is perfectly
managed. (Take the municipal networks, for example, think of 
WiMax, it supports paging).

In addition, there is a performance/security problem with 
paging and dormant mode in general (for which the commercial 
operator does not have a solution). The paging system is easily 
congested. There are two potential reasons:

1/ Flash crowd: Many hosts are called simultaneously. The 
incoming packets trigger link-layer paging. Paging messages, 
hence the IP packets are queued in the system. This is a 
well-known problem. There is no solution. If the paging 
system is congested, you should wait. The call setup delay 
may be very long. 

Nobody tells you that (at the link layer). 

2/ Attack: Attacker (anywhere in the Internet) sends many 
packets to different mobile hosts in the same area. Most of 
the hosts are dormant. Consequently, the above scenario is 
artificially created. The results can be more serious than 
a flash crowd.

You may want to take a look at:
ipmon.sprint.com/publications/uploads/wise06_paging.pdf
(Wireless Security Workshop, with Mobicom'06, Sept. 29, 2006)

In both cases, 1/ and 2/,  the problem is the same:
IP packets don't reach the hosts. They are delayed or lost. 
The loss is *not* caused by increased BER (bit error rate). 
The physical conditions may be perfect. The problem is that 
the link is congested. IP packets are waiting or being lost 
somewhere (between the AR and the hosts). 

The host can detect these problems at the IP layer and employ 
an IP-layer solution: The IP-layer can be told to move to 
another (parallel) subnet. This is the only solution.


> Personally, I'm skeptical. If a node is dormant, why does it want to
> expend resources (like waking up) if the link (at the IP layer) isn't
> working -- given that it is dormant? It's far from clear that a
> dormant node should monitor link status with the same aggressiveness
> it would if it was awake, especially if that has impact on battery
> life.

I'm not sure if the host needs to be aggressive. The test 
interval can be set longer (than active mode NUD). 

Or, the host can check its dormant mode reachability before 
entering dormant mode. 

This has an additional energy cost, yes I agree. This is the
cost of reliability. NUD (standard) also a has a cost. But it 
is necessary. 

pars



> It may well be that it wants to maintain the liveness of the paging
> channel when it is in dormant mode, but my initial reaction would be
> that that would be an L2 issues exclusively, and not something that
> needs to be visible to IP.
> 
> Thomas


--------------------------------------------------------------------
IETF IPv6 working group mailing list
[email protected]
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

Reply via email to