Re: [nbs] NBS and TCP connection identification

[Christian Vogt <[e-mail]>]

Erik wrote:

>> a way to secure IP address changes without the help of Shim6 would be the following:
>> 
>> - Exchange a randomly generated token at the old IP address.
>> 
>> - Verify the peer's knowledge of this token at the new IP address.
>> 
>> - Upon successful verification, associate the peer's new IP address with the peer's DNS name, which happens to encode the IP address that the peer had used during session initiation.
>> 
>> This procedure would be very similar to the Mobile IPv6 care-of address check.
> 
> Explain to me how that can be used to secure the binding between the *name* and an IP address.
> 
> I understand the Mobile-IP approach to secure a binding between two IP addresses, but that isn't the problem that NBS needs to solve.


Erik:

My email above was on verifying an address update, not on verifying a name-to-address binding.  I envision the latter to be secured in one of the following ways:

(1) The recipient operating system verifies the initiator's name by means of a forward DNS lookup.  If and only if the initiator's name is correct, it is delivered to the recipient application.  In an overload situation, the recipient operating system may instead deliver an address-derived DNS name to the recipient application.  This can serve as an identifier for the initiator throughout the session, even across address changes.

(2) Like (1), but with DNSSEC instead of DNS.  Whether to use DNSSEC is a local deployment question.

(3) The recipient operating system always delivers an address-derived DNS name for the initiator to the recipient application.  The recipient application can be provided with an API function to discover the initiator's true name, and have this verified, if so desired.

(4) The recipient operating system always delivers the initiator's name without verification to the recipient application.  The recipient application can be provided with an API function to have the initiator's name verified, if so desired.

What would be your preference among (1) through (4)?  I would be interested in your opinion.

- Christian