After Tymnet had become well deployed in about 1972 a few potential customers came to us with novel security problems. Here are two that I remember.

• Derwent had computerized some large collection of patents dealing with chemicals. They were able to search their data to identify patents describing chemicals meeting some sort of chemistry savvy criteria. Their data base and code were highly proprietary and indeed represented their main proprietary assets. Their potential customer’s queries were nearly as proprietary. Neither party was willing to send their stuff to the other party’s computer for processing. Was there a way for a trusted third party to provide a computing platform that would respect simple rules to protect the respective proprietary interests? See this speculation of followup.
• A consortium of producers and consumers in some market would ask if there was a platform where software agents of consumers could contact software agents of producers to negotiate agreeable terms for purchase. One such consortium was for specialized steel products. There would be many parameters to such a contract and extant negotiating procedures were slow and labor intensive.

About that time Tymshare had become profitable and we could look beyond the current crisis. Three of us, Dale Jordan, Bill Weiher, and myself began a series of Thursday afternoon discussions on how to provide such a platform. Capabilities were known to each of us but none of us assumed that only they held the solution. After a number of these Thursday meetings the design proposals began to take on a pure capability flavor — no other architectures seemed suitable.

In 1972 Dale Jordan wrote a paper outlining the problems and described early elements of an architecture for the SDS Sigma 7. Some time passed and IBM announced a version of the 370 with demand paging. We studied that architecture and deemed it suitable for our plans. The micro processor was still very micro and still years from having the oomph to sport a powerful OS of any variety, let alone oomph to solve our customers problems. Also IBM offered VM/370 somewhat like their previous CP-67 which provided a virtual machine that provided multiple images of a real machine in a decent timesharing environment. At the time Tymshare was providing timesharing services on the SDS 940 and also on DEC’s PDP-10. It seemed clear that a number of customers wanted to use 370 computing remotely. Relatively few companies provided internal interactive access to their main frames. The computing centers of some companies did but were trusted less than a commercial third party such as Tymshare. Tymshare adapted VM/370 to Tymnet and developed a good business selling 370 cycles.

At the same time the virtual machines provided a luxuriant environment to build a new capability kernel which indeed we commenced in 1975.