Codebase Relay

Run ssh -p 2202 base.operand.online and examine our code.

A day ago I discussed a broad plan to rebuild ://operand.online; the challenges are already becoming clear. All the same, some progress!

I focused today on the first motion necessary to upgrade the blog - changing from the clunky share.operand.online to a real git relay, so that someday I can run git clone to copy the markdown to the web client, where I'll rely on SolidJS and MDX to compile the pages.

The essays are in gram:page, the web code is in gram:op, and the deployment code is in gram:build.

So far I'm doing all of the rendering for the essays in Elixir, a fabulous language that has a higher calling than the rendering of markdown.

To change this, I need to first focus on deploying a new relay program, called soft-serve; you can ssh git.charm.sh as an example, or see my deployed copy at: ssh -p 2202 base.operand.online. I recorded a console session as I hooked this up in NixOS.

Now, the relay is up and running. Yay! That is as far as I made it, because the next piece is so uncommon.... that no one really goes that far yet. This is the cool, unknown-frontier piece of coding. We're gonna be in a danger zone from here on.

To be specific about today's errors that are holding me up, I'm seeing different problems locally and in deployment.

Locally, there seems to be a CORS error when I make a connection to https://base.operand.online - an HTTP 405 "Method Not Allowed". I see someone else added code to handle this problem, and somehow it does me no good. Perhaps the issue is that on my local machine, I am running the app on HTTP rather than HTTPS. CORS (cross-origin request security) becomes picky on such conditions.

So, I merged my branch into main and pushed - and the compilation broke because in 2025, compiling to webassembly remains a hard problem somehow. I'm usually amused when I see errors decrying a bad "magic number"; though today I am less than amused - I need this connection.

Now, libgit2.c is a shady binary to compile to webassembly, because the compilation target has been pinned like a donkey-tail to the braying ass of a language.

If I'm going to proceed, I'll be much happier using GitOxide, since much less can go badly when cross-compiling Rust, and I'm beginning to learn a sense of the language tooling.

I'm surprised to say that I'm no longer scared to dig through Rust code - somehow speaking to a bunch of open-source coders who depend on Rust makes the language seem more human, more approachable.

My textbook from No Starch Press has no mention of webassembly, though! I'll be choosing from some capable blog guidance. ( 1, 2, 3, 4, )

Although the blog rendering procedure remains unchanged, I now have a new git codebase relay up and running, happily secured and managed using SSH, and honoring my user SSH keys.

I can now confidently synchronize code again!

Since I spun up the "baseline lab" in Maryland, I've been doing a bunch of on-relay code changes, especially for the code I'm using to clone static.case.law.

Those code changes are soon going to be added to the gram:build codebase, although I'm not looking forward to the merges after a few weeks of divergence between my laptop and the lab.

Sundays are my de facto day off, based on library schedules, and I'm lucky to be busy tomorrow with the Denver Pride fest. I'll sign off here and hope to be back on Monday, sharing anything I picked up around Rust and WASM in the meanwhile.