Summer Minimalism

As the haziness of mid-summer starts to fade away, I am starting to worry a bit more about my interminable personal backlog and getting back to some unfinished projects.

But not just yet. Amidst a series of (perhaps merciful) failures tinkering with generative AI (I’m still quite unhappy with the quality of LLM libraries and overall tooling, let alone the models’ inability to yield useful replies even with retrieval augmented generation and other prompt enrichment approaches), I decided to spend some free time on the minimalist side of computing again.

It happens every Summer, I guess. , revisiting , , , and, last year, .

This year, besides a (which, incidentally, is going well enough to type this) I’m again struck with the notion of finding a to both do quick prototyping and “high performance” services in smaller chips (like low-end ARM CPUs), so I’ve been poking at things like Rampart, Mako and a couple of frameworks.

This would ordinarily be the space I created piku for, except that I’m considering going a bit lower in specs than a modern Raspberry Pi. This means either a stable, zero-hype low-level language or an embeddable scripting one.

But looking back, piku actually came out of a similar exercise, except that one of the design goals was to re-use as much of the typical Linux package ecosystem as possible.

Right now my checklist looks like this:

  • Good HTTP (client and server), JSON and support (i.e., enough “batteries included” to be able to do a lot of things without having to pull in dependencies).
  • Some form of support (because I want it to be self-documenting).
  • Low memory footprint.
  • Good support for parallelism and concurrency, especially concerning networking.

But since I am doing this for fun, I am again going off the beaten path. I grabbed a bunch of related stuff1, ran some of it through and sent the resulting EPUB files to my Kindle to read while on the beach (or, more euphemistically given the current climate, “the ultimate silicon grill”).

  1. I am tempted to use and uLisp, but that might be a tad extreme so I’m focusing on Guile Scheme to begin with since it has been around forever. ↩︎

This page is referenced in: