Re: [RRG] What do we have consensus on?

[Christian Vogt <christian.vogt@nomadiclab.com>]

Lixia, Tony,

I concur with Tony in characterizing Shim6 as performing address
translation.  Internet engineers tend to be skeptical about
translation based on experience with NAT boxes.  But the
(un)acceptableness of translation depends on whether it is stateful,
and if so, whether it takes place in hosts or in the network.

More specifically, in my view, the fundamental difference between
encapsulation and translation is where you store the information
needed to perform a packet transformation.  With classic tunneling,
this information is included in packets.  With classic translation,
the information takes the form of state at the entity performing
packet transformation.  The information carried in packets is then
reduced to a "state lookup key", which enables the receiver to find
the state it needs for the packet transformation.  In classic
translation, the state lookup key is a translated port number.

Accordingly, Shim6 performs translation.  Shim6 completely relies on
state in sender and receiver; the information in packets is limited to
a state lookup key, which in Shim6 terminology is a "context tag".

Classic translation is problematic when performed inside the network,
as NAT boxes do.  The network-side state then constraints the route of
a packet flow.  This limits rerouting flexibility -- a key feature of
IP.  Shim6 does not have this disadvantage because the translation
occurs in end hosts, and hence does not constrain any route.

1-to-1 translation differs from both encapsulation and classic
translation since it requires neither state nor in-packet information.
Due to its statelessness, it does not suffer from aforementioned
rerouting limitations, even if it is performed inside the network.

- Christian



On May 28, 2008, at 10:57, Tony Li wrote:

> The way that I look at it, Shim6 establishes a set of 'identifiers'  
> that the
> transport protocols will use and then plays some games to determine  
> the set
> of 'locators' for network layer operations.  Since it is doing  
> wholesale
> swapping of the locators for the identifiers, I'd claim that it's  
> mostly a
> translation approach.
>
> I think that it's clear that it's not actually changing transport.   
> Further,
> it's pretty clear that while it is piggybacking some information, it  
> really
> isn't going to the full-scale extent of a real encapsulation.
>
> That said, I recall thinking that there was some validity to Brian's
> position and at the same time thinking that it wasn't yet worth  
> worrying
> about the specific details.  I am, of course, very happy to revisit  
> this if
> it becomes productive.
>
> Tony



--
to unsubscribe send a message to rrg-request@psg.com with the
word 'unsubscribe' in a single line as the message text body.
archive: <http://psg.com/lists/rrg/> & ftp://psg.com/pub/lists/rrg