[dns-operations] ip id from servers

[Randy Bush]

>>> There is very little fragmented traffic other than from nameservers.
>> this has been measured?  taken where?  paper cite, plz?
> Just think about all the fragmentation problems with DNSSEC.  These
> would not exist if fragments were common for other traffic.

interesting conjecture.  do tell me when you have demonstrated it.

> Ip id is only 16 bits.  It doesn't take that many replies to have two
> the same, birthday paradox.

and packets from other sources with a similar or larger count of packets
are from the unborn?

on another project, we have been working with ip id for some months.  it
has interesting behavior (e.g. some implementationa properly randomize
while a surprising number still increment).  but we are a bit surprised
by what we think we are seeing here.  so i thougth i would ask.

randu