Re: CGA Use with HBA in Shim6 IETF Meeting July 10, 2006

[marcelo bagnulo braun <marcelo@it.uc3m.es>]

Hi Iljitsch,

El 19/07/2006, a las 15:30, Iljitsch van Beijnum escribió:

> On 19-jul-2006, at 11:57, marcelo bagnulo braun wrote:
>
> > As i understand it, the only way to make the shim6 security based on 
> > IPSec is to assume that a global PKI is deployed, including client 
> > certificates (i.e. not only server certificates) so that it is 
> > possible to secure any-to-any communication.
>
> > From what i understand such global pki is not in place yet and it 
> > doesn't looks like it will be anytime soon if ever. So, i really 
> > don't think it is reasonable to build the security on the shim6 
> > relying on such global pki deployment
>
> > does anybody think that it would be acceptable to build the shim6 
> > security based on the  assumption of a global PKI deployment?
>
> Note that server certificates are relatively widespread, hence my 
> suggestion to adopt TLS as an alternative security mechanism in 
> addition to HBA.


server certificate are more widely used than client certificates 
indeed, but in the case of the shim6 we need certificates for both 
ends, so what do we do for securing the client?

besides, currently deployed certificates provide binding between FQDNs 
and public key.... while in the shim6 we need binding between IP 
addresses and public keys, meaning than currently deployed certificates 
are not good (at least not in the general case and additional 
assumptions about the existance of a FQDN is needed and the associated 
implications...)

in addition, using certificates and public key crypto is much more 
expensive than CGAs, since they would involve public key operations not 
only for the validation of the locator set (as in CGA) but also for the 
validation of the certificates themselves (and this costs grows if the 
certification chain is long). In addition, there is the overhead due to 
the transmission of the certificates in the protocol itself, including 
all the certificates in the cert chain, which may even not fit in a 
single packet so we may end up neededing to send multi-packet messages.

and all this for every shimmed communication....

i thought that one of the key goals in the shim6 design was 
efficiency.... such an approach would really move us apart from the 
efficiency path...

regards, marcelo


> It would be helpful to determine whether we as a wg want this or not.
>
> Obviously it's also possible to use IPsec rather than TLS but I don't 
> see how this would benefit us greatly and IPsec has proven hard to 
> deploy until now.