Re: [Emu] review of draft-ietf-emu-eaptunnel-req-04

["Hoeper Katrin-QWKN37" <[e-mail]>]

See inline.
> -----Original Message-----
> From: Dan Harkins [mailto:[e-mail]]
> Sent: Wednesday, March 03, 2010 3:39 PM
> To: Hoeper Katrin-QWKN37
> Cc: Dan Harkins; Joseph Salowey; [e-mail]
> Subject: RE: [Emu] review of draft-ietf-emu-eaptunnel-req-04
> 
> 
>   Hi Katrin,
> 
> On Wed, March 3, 2010 12:31 pm, Hoeper Katrin-QWKN37 wrote:
> > Dan,
> >
> > OK, I understand that the tunnel provides all these other feats.
> >
> > But why can't the server authenticate during the tunnel protocol? I
> > still don't understand the use case for mutually anonymous tunnels.
> 
>   Because it doesn't have the right credential.
> 
> > If the server has a certificate why can't it send it to the peer
before
> > or during the tunnel establishment?
> 
>   If the server has a certificate then sending it to the peer
> would not really solve any problem. The peer would still need to
> have a reason to trust it and we're back to the problem of putting
> a trusted certificate in some certificate store. A global PKI to
> solve all of our certificate issues still has not materialized.
> 
> > If the peer and server share a secret, than this could be used to
> > establish the tunnel.
> 
>   If the peer and server share a secret they could use one of the PSK
> ciphersuites for TLS but those are susceptible to a dictionary attack
> and are therefore inappropriate.
> 
>   The tunnel is being established with EAP-TLS so we are limited to
> TLS ciphersuites and the authentication they provide. If a TLS
ciphersuite
> was appropriate always and everywhere then we would not need any other
> EAP methods, we'd just do EAP-TLS. But that is not the case. Also it
is
> a requirement to tunnel additional EAP methods inside the tunnel so
> obviously there are EAP methods that provide something that a TLS
> ciphersuite does not.
> 
> > What I am saying is what kind of server authentication credentials
could
> > be used inside an anonymous tunnel that could not be used to
> > authenticate the server in the tunnel protocol? (given that privacy
is
> > not the issue)
> 
>   A low-entropy password that can easily be remembered and entered by
a
> human with low probability of error.
[KH] I asked what kind of SERVER credentials not peer credentials. 
> 
>   Dan.
>