Re: [hiaps] Other Built-in Demux Identifiers: The QUIC Approach

On Wed, Nov 12, 2014 at 9:28 AM,  <mohamed.boucadair@orange.com> wrote:
> Hi all,
>
>
>
> FWIW, QUIC is proposing a unique identifier to demux connections bound to
> the same IP address.
>
>
>
> “Given that NAT services can vary the WAN-visible source-port over the
> lifetime of a connection (via unbinding, and rebinding), it is clear that
> source-IP address and source-port are strictly insufficient to define a
> connection. We overcome any such confusion by using a CID (Connection
> IDentifier) which generally persists for the lifetime of the connection.  A
> CID is a pseudo randomly generated nonce [size is currently set at 64 bits],
> that is expected to be universally unique.  The CID is usually proposed in
> the first UDP packet sent by a client to a server, and is present explicitly
> or implicitly in all future packets that are exchanged for the lifetime of
> the connection.  It is the defining key for the connection.
>
>
>
> A server can use this CID as a key to identify the specific connection from
> among many inbound connections sending UDP packets to a single server port.
> The CID is also used by the server to resurrect the current session
> encryption context for use with AEAD (Authenticated Encryption with
> Associated Data).  That context conceptually includes an encryption key, as
> well as an authentication key.  (Note: mutable portions of the packet, such
> as source-port, etc., are not included in the Associated Data
> authentication).”
>
>
>
> More information can be found here:
> https://docs.google.com/document/d/1RNHkx_VvKWyWg6Lr8SZ-saqsQx7rFV-ev2jRFUoVD34/mobilebasic?pli=1
>
>
>
> Note CID is universally unique.
>

Yes but QUIC is designed as connection identifier and as such it does
not seem to be appropriate to use it as host identifier.

Regards,

Behcet
>
>
> Cheers,
>
> Med
>
>
> _______________________________________________
> hiaps mailing list
> hiaps@ietf.org
> https://www.ietf.org/mailman/listinfo/hiaps
>