Caterpillar

I've talked about the possibility of exporting a program from the interactive runtime, into a fully compiled and optimized form for production deployment. But what if we had a more advanced, more flexible form of that concept? What if we could decide on the fly which parts of the program are "solid", i.e. compiled and optimized, or "fluid", i.e. with full support for interactive programming.

At any point, we could decide to change the boundary between solid and fluid, whether the program is running in a local development environment or a remote staging/production environment. That might take a few seconds, compiling/optimizing/linking the solid code and deploying the new version, but so what? Afterwards, we could manipulate the fluid parts of the program in a fully interactive way.

This is just an idea. Maybe it wouldn't be as useful as I imagine it. But if it worked out, it would turn interactive programming in Caterpillar into a zero-overhead abstraction. It would no longer slow down code that doesn't use it.