[cap-talk] network level designation and authorization

[David Hopwood]

coderman wrote:
> On 6/8/06, David Hopwood <david.nospam.hopwood at blueyonder.co.uk> wrote:
> 
>> ...
>> This is so different from how a VPN is normally used that I would suggest
>> not describing it as a "VPN", even if the intention is to use crypto
>> protocols that are most commonly used to implement VPNs.
> 
> fair enough; how would you describe it?

As a secure session protocol.

>> There isn't any particular commitment to specific crypto protocols like
>> TLS in the high-level design of the web-calculus or CapTP/VatTP, if that
>> is what you are worried about.
> 
> this was the concern.
> 
> providing communication privacy below the service level allows this
> layer to be updated or modified according to various needs (for
> example, the difficulty upgrading digests in TLS), while the
> functionality and security of the capability services is isolated from
> these headaches.
> 
> regarding CapTP: are the VatID's thus entirely separate from
> communication privacy?

The VatID includes a cryptographic hash of the vat's long-term public key.
This would not prevent the long-term public key from being used to sign
ephemeral public keys needed for different protocols.

> it appears this is the case but the mention of
> CapTP for privacy confused me.

My understanding is that the CapTP design could work over any secure session
protocol.

> also, in the CatTP parameter descriptions[1] i see mention that a
> nonce "May be guessable, but must not accidentally collide."
> 
> i have always assumed that the unguessability of nonces was central to
> the security of the capabilities / resources they are used to
> represent (and also why i am so fond of truly random number generators
> in hardware).

I'm not sufficiently familiar with the details of CapTP to answer this.
However, it's not uncommon for security protocols to require nonces that
are unique, but do not have to be unguessable.

-- 
David Hopwood <david.nospam.hopwood at blueyonder.co.uk>