[comp.protocols.tcp-ip] Network connectivity

CLIVE@XX.LCS.MIT.EDU (Clive B. Dawson) (11/04/88)

MCC.COM (10.3.0.62) has unable to establish connections with a large
percentage of the Internet hosts during the past 24 hours.  I believe
much of this is caused by the fact that a lot of gateways have been
turned off to try and prevent the spread of the Unix virus that's been
going around.  Furthermore it looks like there's a lot of "one-way"
gateways out there.  OUr SMTP listener has had a lot of connections
hanging in SYN.SYN state and timing out, presumably because our SYN's
aren't making it back to the host who is trying to connect to us.

But I really got confused when I discovered that many of the hosts
I couldn't connect to were reachable from other hosts on the 10 net
(e.g. Score and XX).  If anybody can offer one or more explanations for
this, I would very much like to hear about it.  In particular,
what are the possible causes for a host being reachable from 10.a.b.c
but not from 10.x.y.z?  (Note that MCC.COM's "default,preferred"
address IS 10.3.0.62.)  The only explanation I can come up with 
is that something is messed up in the PSN routing tables, and that
this is a problem for the NOC.

Thanks,

Clive
-------

goldstei@MITRE.ARPA (Steve Goldstein) (11/04/88)

Clive,

I have had similar experiences a few weeks ago with a Princeton host on
arpanet and reachable thru nsfnet via jvnc.  The telco line from jvnc to pri
on broke, and there was no way in hell that I could get through to princeton
via my normal methods.  Then, I logged in to a machine in Wisconsin that
happened to be on arpanet and, voila!  there I was getting into the princeton
host!  Seems that the new routes did not age properly on jvnc's EGP-server.
They should have shown the new route to princeton via arpanet, but they did not.

I have no idea in this current swiss cheese Internet whether or not similar
things could be happening, but certainly the gated, EGP and other things that
usually hum must be going bump in the night!

Keep the faith,

Steve Goldstein
normally: goldstein@ames.arc.nasa.gov (which is down!)