Certifiable Minicomputer Project and ARPANET

For Don Good, Woody Bledsoe, Jim Browne, and the Institute for Computing Science and Computer Applications, the DEC-10 offered a large address space and stable multitasking capabilities. Their work aimed to treat computer programs as formal mathematical objects, completely replacing empirical, trial-and-error debugging with mathematical certainty. ICSCA was located on the 20th and 21st floors of the UT Tower. In the seventies and early eighties, the major research thrust of ICSCA was program verification. Motivated by deep concerns over the unreliability of software controlling critical infrastructure and nuclear armaments, Good sought to replace trial-and-error debugging with rigorous mathematical proofs. To achieve this, the team developed the Gypsy Verification Environment, an interactive system running on the DEC-10 that transformed programs and specifications into logical formulas to be mechanically proven. 

In the Certifiable Minicomputer Project of the late seventies, the UT team used the DEC-10 and Gypsy to achieve a milestone for the era by mathematically verifying ARPANET security code associated with historically important encryption hardware from Bolt Beranek and Newman. Developed by BBN starting in 1973, the Private Line Interface aimed to give ARPANET users the equivalent of a private, leased line over a shared public infrastructure. The PLI pioneered the concept of edge cryptography for packet-switched networks, establishing an architectural trajectory that eventually evolved into today's IPsec and High Assurance Internet Protocol Encryptor standards. 

To operate at the edge of the network without requiring extensive modifications to the existing ARPANET infrastructure, the PLI relied on a breakthrough known as selective payload encryption. By encrypting only the sensitive message contents while leaving the machine-readable routing and addressing headers in plaintext, the PLI allowed classified traffic to be securely routed through shared IMPs without requiring the internal network switches to hold any cryptographic keys. The devices were approved by the NSA in 1975 for limited deployment to protect classified data and originally utilized manually keyed cryptographic units. The success of the PLI's edge-based architecture quickly spawned a lineage of advanced successors that shaped the modern internet. 

The physical, single-purpose black box edge devices engineered by BBN inspired UT Austin's Certifiable Minicomputer Project. Because dedicated hardware like the PLI was rigid, defense agencies wanted to know if a more flexible, software-defined gateway running on a minicomputer could provide the exact same level of edge security. By successfully writing and mathematically verifying a 5,000-line cryptographic gateway in the Gypsy programming language, the UT Austin team proved that software could sit at the edge of the network and securely interoperate with BBN's physical hardware without leaking data, proving that absolute security guarantees could be enforced in software at the network's edge. As described in the quote from Clive Dawson, a PLI was installed and required that a very thick cable with 32 twisted pairs of copper be run from the 21st floor of the tower, down the elevator shaft and through the steam tunnels to the Computation Center, where it plugged into the IMP. When the CMP project was completed several years later and the hardware decommissioned, the cable was left in the elevator shaft, as it was not worth the trouble to remove it. [1]

South entrance to the Computation Center. The cable route was through the steam tunnels under the terrace and the Main Building, then up the Tower's elevator shaft. Sad to note that this was also essentially the response route used on August 1, 1966, when the Computation Center and the CDC 6600 were young.

[1] Here is a full quote thanks to Clive Dawson. This concerns the second IMP at UT, following on from the 1977 first IMP that had initially been located in HRC with the DEC-10. Sometime around 1978-80, a newer IMP was delivered by BBN and installed in the Computation Center. A PDP-11/70 was installed there to connect to the new IMP, and so the UTEXAS host was rehomed from TSP to the Comp. Center. It was only when the first DEC-20 arrived and was installed in Painter Hall that the PDP-11 became UTEXAS-11 and the 20 was named UTEXAS-20. As I recall, the 11 retained the host nickname of UTEXAS. At about the same time, specialized research hardware was acquired by the Certifiable Minicomputer Project (CMP) for the purpose of providing secure encrypted communications. This required that a very thick cable with 32 twisted pairs of copper be run from the 21st floor of the tower, down the elevator shaft and through the steam tunnels to the Comp. Center, where it plugged into the IMP. When the CMP project was completed several years later and the hardware decommissioned, the cable was left in the elevator shaft, as it was not worth the trouble to remove it. I wonder if the folks currently remodeling the Tower have come across it?! (Hat tip to Rich Cohen and Clyde Hoover for filling in some details of this story.)