Re: 6man w.g. last call for <draft-ietf-6man-default-iids-11.txt>

[Lorenzo Colitti <[e-mail]>]

On Sat, May 14, 2016 at 11:32 AM, Mark Smith <[e-mail]> wrote:

> The thing wrong with "solving" a layer 3 problem at layer 2 is that it
> makes the effectiveness of the layer 3 outcome entirely dependent on the
> attributes of another layer. It's fundamentally a layer violation.
>
> I think we have layers to try to minimise or avoid these dependencies.
> Having a layer 3 privacy solution to this layer 3 privacy problem means
> that it doesn't matter if the layer 2 privacy is terrible or not.
>
Actually, it's the opposite: the reason we have layers is so that layer X
doesn't have to worry about reimplementing functionality provided by layer
X-1.

If we were to follow your argument, we would implement CRC checks, packet
retransmissions, and fragmentation into HTTP, because otherwise the
reliability of HTTP would be entirely dependent on the attributes of layers
2, 3, and 4. Guess what: the outcome of HTTP *is* entirely dependent on the
link-layer not scrambling data, on on TCP dealing with packet reordering
and packet loss by sequencing and retransmitting, etc.