[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: Proposed BOF charter and agenda



Hi Steve, 

> -----Original Message-----
> From: owner-ietf-ipsec-failover@xxxxxxxxxxxxx 
> [mailto:owner-ietf-ipsec-failover@xxxxxxxxxxxxx] On Behalf Of 
> Stephen Kent
> Sent: Monday, February 05, 2007 8:30 AM
> To: Paul Hoffman
> Cc: ietf-ipsec-failover@xxxxxxxx
> Subject: Re: Proposed BOF charter and agenda
> 
> 
> At 2:22 PM -0800 2/2/07, Paul Hoffman wrote:
> >...
> >The purpose of this working group is to define necessary 
> payloads to support:
> >
> >1) Negotiation of failover recovery capability
> >2) Server to client state transfer for stateless recovery
> >3) Client-gateway IKEv2 session resumption
> >4) IKEv2/IPsec state and corresponding format needed for recovery
> >
> >Support for capabilities beyond those listed above is out of scope: 
> >more precisely, specification of a gateway to gateway state 
> transport 
> >protocol, protocol or payload extensions or modifications to support 
> >load balancing between gateways is out of scope.
> >
> >=============
> >
> >Note the last paragraph. In Russ' preliminary discussion 
> with the IESG 
> >and IAB, there was a lot of concern that a server-to-server 
> part would 
> >result in a serious rathole.
> 
> Paul,
> 
> Given the argument cited above about not wanting address 
> server-to-server communications, I am surprised by #2 above, 
> since the use of server-to-client state transfer seems to 
> suggest that had to be a server-to-server transfer of the 
> state!  If #2 were client-to-server state transfer, then this 
> issue would not arise.
> 
> It's not that I personally prefer one over the other, but 
> rather that I don't want to see us adopt an approach that 
> implicitly calls for a capability that we explicitly have 
> chosen to NOT standardize.
> 

#2 is referring to the case where the initial gateway is providing the
state to the client that can be presented by the client to a new gateway
upon failover. "Stateless" there refers to the backend operation - when
the state is stored in the client, the infrastructure can remain
stateless for the purpose of failover (i.e., no state needed on backup
gateways). Re-reading the charter text, it looks like we have no
explanation of what we mean by stateful or stateless in that text.
Perhaps, we should clarify that to avoid confusion. 

Would that address your concern? 

Vidya

> Steve
> 
>