>>> (3) Remove the I-DPD mode from IPv4 and only have a hash mode. A
>>> non-assisted hash mode would of course not be robust since packets are
>>> sometimes produced that have the same content. Not a fan of this
>>> although its an easy change.
>> This is already the case, FWIW, for IPv4 packets with current ID use. The IDs wrap fast enough that you already see packets with duplicate IDs that are NOT duplicate packets.
> Ok, that is a way forward. Go with non-fragment/non-IPsec hash mode
> and also keep the more explicit IPsec, fragment processing discussion
> that I believe is ok.
I think that's what's already in the IP-ID doc at this point, so that'd
be good (I don't go into detail with the IPsec variant, but that's not
critical IMO to the IP-ID doc).