[CGA-EXT] comments on draft-jiang-csi-dhcpv6-cga-ps-03.txt

[marcelo bagnulo braun <[e-mail]>]

Hi,

i have a few comments on  draft-jiang-csi-dhcpv6-cga-ps-03.txt

In section 3. What DHCPv6 can do for CGA it reads:

   Generating a key pair, which will be used to generate a CGA, also
   requires a notable computation. Generation and distribution of a key
   pair can also be done by DHCPv6 server. Of course, when designing
   these new functions, one should carefully consider the impact on
   security.  However, the security considerations of specific solutions
   are out of scope of this document.


While i agree that the security aspects of a specific solution are  out 
of scope, i am not sure we can completelly dump the issue. I mean, in 
order for the DHCP server to convey the SEC information, the ecurity is 
critical. Is it really feasible to provide enough security, without 
breaking the dhcp model? I think further analysis on this is needed.

Then, in 4. What CGA can do for DHCPv6, it is described that  CGa can be 
used to secure dhcp. Now, i think a bit more analysis of what features 
would be provided if we do this i.e. what types of attacks are 
prevented, it would be useful AFAICT, this would much like an ssh type f 
security (i.e. also called oportunistic or leap of faith security) I 
think this is worht the trouble, But i think needs to be more clearly 
stated.

So, if we cover these two topics, i think the document does a fairly 
good job analysis the different apsects. Now, i think it would be 
interesting also to discuss (even though maybe not include in the 
document at this point) what parts of this interaction we would like to 
work on if any.

Regards, marcelo