I was their Corporate Webmaster in the 1990s (effectively CTO, but not quite). The networking support teams had this in place back then. If you did not say you

I was their Corporate Webmaster in the 1990s (effectively CTO, but not quite). The networking support teams had this in place back then. If you did not say you wanted bi–directional traffic, then you would not get bi–directional traffic. They were incredibly dependent on their filter rules, and their process assumed the filter rules were always correct and accurate. One night I got confused about which window I was in and ssh'd into my desktop machine (on the internal 9/8 network) from one of the ibm.com backend servers (on the open internet). A "routine update" caused routes for 9.0.0.0/8 to be advertised on the open internet. Although I had a short escalation path it still took an hour from when I discovered the hole to get anyone in Global Services to accept it as a problem.