Re: [dhcwg] Call for comments - LDRA

[Ted Lemon <[e-mail]>]

On Sep 30, 2009, at 7:58 PM, MILES DAVID wrote:
> 	 is a valid DHCPv6 message (per RFC 3315)

Is there a reason for this exception?   I would argue that it should  
intercept all packets that meet the other qualifications, including  
packets that turn out not to be valid.   If they are not valid, it  
drops them on the floor.   What would be the benefit in forwarding them?

> The LDRA copies the peer-address into to the destination IP address
> field and SHOULD forward the packet to the correct client-facing
> interface using the destination link-layer (MAC) address or the
> Interface-Id in the Relay-Reply.
>
> Do we go for a MUST?

Two things.   First, I think a MUST is unnecessary to ensure  
interoperability.   Second, I think having one clause of a  
specification sentence that doesn't have a SHOULD and another that  
does is a bad thing.   So I would further reword this:

The LDRA copies the peer-address into to the destination IP address  
field.   The LDRA SHOULD forward the packet to the correct client- 
facing interface using the destination link-layer (MAC) address or the  
Interface-Id in the Relay-Reply.  The LDRA SHOULD NOT retransmit the  
packet on any other interface.

The reason I say to not make it a must is that you may have a case,  
probably _not_ in a cable modem or DSL environment, where the device  
isn't smart enough to do this, and I don't think it's necessary to  
make such a device non-conforming.

> Thanks again Ted,

Thanks for working on this!