Re: [DNA-BOF] Do we need a separate DAD Problem Statement? (was Re:About Charter, DAD, and SEND)

[Bernard Aboba <aboba@internaut.com>]

> It SHOULD be possible to optimistically use optimized
> DAD addresses even before the DNA process has not
> been completed.  In other words, the hosts SHOULD be
> allowed to send packets to a newly attached link with
> the optimistic assumption that the link is the same
> one that was previously used, without needing to wait for
> the DAD and/or DNA procedures to complete.

I think this requirement is ok.  It fits several of
the assumptions already made in the DNA documents, including those
relating to the role of ARP/ND in DNA.

In the absence of other information, the DNAv4 document states that the
host is to assume that it remains attached to the previous point of
attachment.  If the host has not changed subnets, then ARP/ND for the
default gateway will succeed, and this validates that the host is still
present on the assumed subnet, allowing it to keep using a valid IP
address.  The only additional wrinkle in DNA is that it is required that
the ARP/ND demonstrate that the default gateway has not changed MAC
addresses.

If the "optimistic" traffic purely local, and the ARP/ND cache entry
hasn't aged out yet, then it is possible that the host will be putting
traffic onto the local link which isn't being received because the
destination is actually off-link.  In this case, losses will continue
until DNA completes and MIP handover occurs.  I'm not clear that this
behavior is as "optimistic" as might appear at first blush, since the
result is a backoff of RTO timers and a change in TCP state that might not
occur if the host were not so "optimistic".

For example, if the ARP/ND cache had aged out, and the host has changed
subnets, yet is still using an address inappropriate on the new link,
then presumably ARP/ND will fail, since the default gateway and/or
destination host is not present on the link.  On detection of the
failure, it is appropriate to obtain an alternate IP address.