On the Design of Display Processors (1968) [pdf]

I absolutely love the original analysis (which is why I shared it).

The hardware version: anything capable of general purpose work is too important to use- so you delegate to specialized hardware. That hardware then evolves the capability to do general work and therefore becomes itself too valuable to use.

The software version: Domain Specific Languages, Remote Procedure Call, and many others.

Current graphics ideas: shaders, CUDA, and so on.

It's interesting how something like the opposite ended up happening. Microprocessors became so cheap that they're used in almost everything. In theory, your keyboard and monitor can do general work, if you update their firmware to tell them what to do. Even my (electric) toothbrush contains a microprocessor.

The conclusion is amusing in retrospect (obviously things have changed since then!):

"The view suggested by Daniel Bobrow that the display processor need not, indeed should not, contain mere general purpose computing power, largely determined the design of our display processor."

Some days it seems the best parts of the computer revolution are over; other days it seems it's only just begun.