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

["Bound, Jim" <Jim.Bound@hp.com>]

IPsec is deployed end-to-end for v4 and v6 in production not sure I
agree no one knows how to do this and I think I misunderstood your
statement below?  Thanks.
/jim 

> -----Original Message-----
> From: Iljitsch van Beijnum [mailto:iljitsch@muada.com] 
> Sent: Tuesday, July 11, 2006 9:09 AM
> To: Pekka Savola
> Cc: Bound, Jim; shim6@psg.com
> Subject: Re: CGA Use with HBA in Shim6 IETF Meeting July 10, 2006
> 
> On 11-jul-2006, at 8:01, Pekka Savola wrote:
> 
> >> Recommendation: For now remove HBA and the use of ULID security 
> >> specifying HBA and leave it as work to be completed that 
> avoids this 
> >> IPR problem with CGA.  Suggestion is to simply embed ULIDs 
> within the 
> >> data payload with new option and secure all communications 
> at least 
> >> for now for IP layer communcatiions with IPsec encryption based on 
> >> locator pair.
> 
> > How could this any-to-any IPsec (no prior relation to your 
> peers can 
> > be assumed) be made to work?`
> 
> > I think this potential solution path was hinted at the security 
> > directorate review we got some time ago,
> 
> And what would that be?
> 
> Let me repeat my suggestion from yesterday in a bit more detail:
> 
> In the absense of any information on whether there is even a 
> patent claim on HBA itself or only on CGA (is there really no 
> way to determine this?) we keep HBA but rather than make it 
> the only and mandatory security mechanism (I'm unsure about 
> the status for CGA but I'll assume that to be a variation of 
> HBA for these purposes), we add an additional security mechanism.
> 
> What I'm thinking of here is not IPsec, as nobody has yet 
> figured out how to run IPsec between two hosts without any 
> prior coordination in the real world, but rather TLS. This 
> could be as simple as doing the shim negotiation inside a TCP 
> session encrypted with TLS. This means more and larger 
> packets, and one end needs to have a certificate that leads 
> back to something the other end trusts, so in some ways it's 
> inferior to HBA, but TLS is widely implemented so adding this 
> should be very simple and then people have a choice between 
> HBA IPR issues or TLS PKI and performance issues. Last but 
> not least, TLS security would make it a good deal easier to 
> implement the shim in a middlebox.
> 
> Iljitsch
>