On Thu, Nov 19, 2009 at 2:35 PM, Dae Young KIM <[email protected]> wrote:
> On Thu, Nov 19, 2009 at 6:23 PM, Patrick Frejborg <[email protected]>
> wrote:
>
>>
>> >  And only one identifier.
>> >
>>
>> Only one...?
>>
>> IMHO, a PKI certificate identifies a stack/person/host so it is a
>> identifier in the RRG terminology, right?
>> http://trac.tools.ietf.org/group/irtf/trac/wiki/RRGTerminology
>>
>> A second identifier is needed, that will provide mobility (fixed and
>> mobile site, endpoint) and not as complex to deploy as a PKI
>> infrastructure, also less secure than the PKI infrastructure. Think
>> this needs be clarified, if not - there is a risk that the new
>> identifier will have too much security features and start to compete
>> with the PKI infrastructure??
>
> When moving, your identifier will still be kept, not changed, but your
> locator will be changed. The mapping between the identifier and the changing
> locator (and its retrieval) will have to be done a server in the
> infrastructure (perhaps an extended DNS or rendezvous server(?) in HIT) a
> very efficient manner.
>

Yes, this is the host identifier approach - very similar as a PKI
infrastructure but the PKI doesn't offer mobility, the PKI certificate
do have global uniqueness . Another approach is to use a session
identifier that can offer mobility, the session identifier is not
globally unique - it is just used to identify the session when the
endpoint is moving from one attachment point to another.

Both identifiers can be used concurrently, if the context of the
session is sensitive then use the session identifier for mobility and
to identify the remote endpoint after the transition authenticate
again the endpoints by the PKI infrastructure.

If the content is not sensitive - do you need to authenticate the
remote endpoint again? Probably not, but mobility might be required
and for that the session identifier is good enough - usually it is the
client that moves around and the server is fixed. Or if both endpoints
moves around you would need a rendezvous server - but I think has been
solved on the application layer already, e.g. SIP registrar&proxy,
Skype, Instant Message solutions, peer-to-peer applications etc.

The problem is that some applications uses IP addresses to identify
the session, because there is no session layer in the TCP/IP-model
(the OSI model and Appletalk do have) - the lack of the session layer
is in my opinion the problem. So if the application could identify the
session with the help of a token much better mobility could be
achieved.
Unfortunately it would require changes to some applications - but it
is the right place to fix the problem. If the application can not be
changed and mobility is required, then use Mobile IP.
A simple session layer would make the networking so much easier.

-- patte
_______________________________________________
rrg mailing list
[email protected]
http://www.irtf.org/mailman/listinfo/rrg

Reply via email to