The work, with marginalia.

Peter Deutsch and I were trying to make a wonderfully interactive language run like a practical product. We called the technique dynamic translation. The world later settled on “JIT,” which is punchier and, I still think, less accurate.

A prospectus is written before anyone knows which parts of the future will arrive on schedule. This one belongs to my MCC years, when “enterprise integration” still needed a fairly elaborate argument.

The title is pure 1992, but the problem is still recognizable: make institutions, suppliers, and their incompatible systems move at something closer to the speed of the need.

The Web could display a catalog; the harder question was whether strangers and companies could actually do business there. “Spontaneous commerce” was our name for the missing machinery.

We carried lessons from electronic commerce into health care: heterogeneous institutions, documents that outlive systems, and no single authority capable of ordering everyone onto one platform.

This began with a mildly unfashionable question: what if the links around a page tell you more about similarity than the words on it? The result was a useful reminder that structure is often data wearing work clothes.

Personalized services generally ask for precision they do not need. We explored the opposite bargain: reveal a less exact version of yourself and see how much utility survives. Quite a lot, as it happens.

People attach privacy expectations to messages all the time, usually in prose and usually without leverage. RAPPD asked what happens if those expectations become explicit enough for software to notice.

The interesting sites here were not the famous ones. We looked at ordinary sites without privacy departments and found that accidental disclosure is still disclosure—even when nobody intended to build a surveillance machine.