In your previous mail you wrote:

   About section 2 on Correspondent Nodes; could you elaborate in the 
   document why exactly solution is too drastic?

=> because this gives no choice between bidirectional tunnel and
route optimization, so in some cases mobile IPv6 becomes far less
attractive. The real impact depends on how mobile IPv6 is used,
in fact one can argue that bidirectional tunnels are enough, but
I don't believe that mobile-ip list members will agree...

   Note that BCE check is not 
   the only way to ensure legitimity of HAO: if it's secured by AH, it's ok;  
   if some SUCV/.. weak authentication method is used, it's probably also ok; 
   the same might even apply to return routability.  It's too early to crush 
   CN solutions.
   
=> these CN solutions have the same cost than full routing optimization,
so I consider them as BCE check variants.

   (I think the solution for HAO should most likely consist of two separate, 
   "strong-enough" layers, one mandated at CN, one possible at firewalls, but 
   that's not the topic of this draft).
   
=> one mandated at CN == no third choice.

   Note: it seems every site, even if it had only a few MN's, will have to
   have AAA infrastructure, so that it could interact, certify etc. home
   address use for remote AAA systems when MN goes roaming and there's a need
   to punch a hole in ingress filtering of remote sites.
   
=> I don't know if the "remote sites" are home sites (sites with home agents)
or correspondent sites (sites with correspondent nodes).
 In the last case the only issue is the iDDoS because both care-of
and home addresses are external. In the first case one can rely on
home registrations (which have to be strongly secured) in order to
understand what happens (and what HAO are valid), in fact this is just
remote network access control (in AAA terms this can be done directly
(IKE with certificate for instance) or (better because simpler) using
the local/visited AAA system as a mediator, note the issue is for
first home registrations because they have to create security contexts).

   (If this is the approach for security, it should be required in the main
   MIPv6 draft).
   
=> I disagree, the iDDoS security threat is not a major one because
ingress filtering is not really mandatory. The purpose of my draft is
not to fill the hole, it is to get back the previous (i.e. before HAO)
situation.

   Or have I missed something?  This seems unnecessary in many environments,
   e.g. university campus area WLAN or company's internal network.
   
=> this depends on where are the visited, home and correspondent domains.
 For an university campus area WLAN home agents are in campus area too
so a bidirectional tunnel with a good local security seems enough
(i.e. visited domain = home domain with home agents in the path
between MNs and CNs).
 In a company internal network all three domains are the same so there is
no security problems as soon as basic security (i.e. no physical intruders)
is enforced.

Regards

[EMAIL PROTECTED]
--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page:                      http://playground.sun.com/ipng
FTP archive:                      ftp://playground.sun.com/pub/ipng
Direct all administrative requests to [EMAIL PROTECTED]
--------------------------------------------------------------------

Reply via email to