Flipping the Bozo Bit
Keith Irwin & Christoph Neumann

Episode 7: Steel Mills to Data Warehouses

2013-05-19

What if you can’t duplicate your production stack for testing and so on? For instance, factory software? I’m starting to think this whole podcast is, ultimately, about design.

In This Episode:

  • Steel mills
  • Stackless programming (and we ain’t talkin’ ’bout push/pop).
  • What if you can’t run your software except in production?
  • You can test but you can’t run.
  • Your interfaces can never be too thin.
  • Adapters outside the walled garden of perfection.
  • The immutable log of perfection.
  • What’s the difference between data streaming in from hardware, or replayed log data?
  • The wall as observation post.
  • You gotta go batch.
  • The golden nugget and the matrix.
  • Everybody gets a copy!
  • How to engineer bugs for performance review goodness.
  • “This whole things screams, ‘Messaging! Messaging!’”
  • “This Facebook thing is confusing me a bit.”
  • How many gamma bursts per angstrom?
  • Human vs machine interface and the idea of a command protocol expressing user intentionality. Beep boop.
  • Human typeable protocol, the ultimate human/machine interface. (Well, outside of brain implants. Maybe.)
  • The Hardware shouldn’t know about the System, the System shouldn’t know about the Hardware. (Or Facebook.)
  • If software devs and hardware devs agree on a simple protocol they can fake the rest.
  • Any time you can remove dependencies from the system, it’s good.
  • Hardware interfaces are one thing, but why not do the same thing with remote software interfaces? (Say, Facebook!)
  • Facebook is hard. Ware.
  • The footzy-boo camera, not that old whatever-whatever camera.
  • System integration should be about intentions, not implementation details.
  • Distributed adapters mean incremental upgrade. No down time!
  • Erlang pretty much got all this right long ago.
  • Everyone needs to learn to walk, even though “walking” has been solved.
  • That’s right. Steel mills to data warehouses.
  • We don’t build one-off lamps, we build lamps that morph and evolve.
  • It’s about the data that flows, not the nodes it flows through.
  • Does it come down to separation of expertise as well as concerns?
  • “I thought we were experts on everything.”