From:	MERC::"uunet!CRVAX.SRI.COM!RELAY-INFO-VAX" 21-OCT-1992 19:17:49.18
To:	info-vax@kl.sri.com
CC:	
Subj:	RE: What's the right OS for a packet switch?


	 ... why Unix, of course.  Packet switches for the major backbone
	network of the United States don't need to be reliable, they don't
	need to be flexible, they need to be Unix...  Herewith the proof, an
	excerpts from a recent trouble ticket ...

	    [Original report:] On the ENSS rcp_routed has crashed:
	    Oct 20 00:58:55 enss134 syslog: rcp_routed must be restarted
	    Oct 20 00:58:55 enss134 syslog: rcp_routed must be restarted

	Of course having detected that a vital system daemon needs restarting,
	our good little Unix just lets you know rather than doing it...  And
	why did it need restarting?

	    [The entire T3 NSFnet backbone went unusable for an hour] Merit
	    believes the problem tonight was related to ESNET trying to come
	    up on the T3 backbone-- they reached a threshold.

	Right.  A fixed length array overflowed and the packet switch crashed.
	I guess I didn't really want reliable networking...

No, no, no.  You don't understand how radio works.

Look, all the machines connected to the network are Unix machines, right?
(If they are not, they soon will be.  Deviationism will NOT be tolerated.)
Those machines couldn't take the full speed, full-reliability connections
anyway.

What you are seeing is NOT a bug.  It's a form of flow control.  (Oh, and you
probably didn't recall that such error messages are usually typed at the
console by code running with interrupts disabled.  See, flow control again.)

							-- Jerry