the facts are, internic's own servers (f.gtld-server.net etc) were sending
back NOERROR/NOANSWER whereas my server was sending back a lame delegation.
that means clients would try other servers to gain the right final answer
if they got a bad answer from my server, but they would just report a fatal
application error ("host not found") if they contacted an internic gtld
server.
the other fact of note is that the reason my server was lame had to do with
some kind of provisioning problem inside internic or its isp. my own mrtg
graphs are clean for this period, and internic was the only site i couldn't
reach at high bandwidth. f.root-servers.net was lame for two days because
it could not fetch COM from NSI.
yet NSI, in its statements to reporters and to users, are of the following
form. it is REALLY DAMNED HARD to not get pissed off at this kind of BS.
*MY* server was not sending fatally bad answers, *THEIRS* were.
i've been in touch with wired news and have been queried about this from
CNN and i'll be telling the whole story to every reporter who can reach me.
i've seen holtzman's comments about "unreliable volunteer servers" in the
WSJ in the past and i'm no longer willing to sit by and be slandered when
it was NSI's problem all along. (tcpdump doesn't lie.)
(a total of five servers were having trouble, of two different kinds, so
this was a complex failure situation.)
------- Forwarded Message