By Jugert Mucoimaj · Tirana, Albania
Engineering notes
Two implementation notes from public projects. They describe the problem each tool addresses, the design visible in its repository, and the tradeoffs worth examining in the source.
Last updated:
envless · Python / FastAPI / PostgreSQL
Rotating configuration while applications are running
A service that reads a secret only at startup cannot use a replacement until it restarts. That is inconvenient for routine configuration and risky when a credential needs to be retired. envless keeps application clients connected to a self-hosted configuration service. After a value changes, clients receive an event and re-read their configuration. The repository documents this as a live update path built around a stream endpoint rather than a deploy step.
Why the stream signals a re-read
The example client listens to /v1/stream, then fetches current configuration when an event arrives. That separates the notification from the value itself. A client that missed an event can still ask for the current state; the event is a prompt to reconcile, not the source of truth. The example also shows a client pulling configuration at boot and continuing to run if the service becomes temporarily unavailable. Those behaviours are visible in the public client code and are more useful to inspect than a claim that updates are simply “real-time.”
What a rotation must make visible
Changing a secret is only part of rotation. An operator needs to know which applications have caught up. envless versions writes, records audit entries and displays the progress of connected applications after a value changes. Its README also documents encrypted values at rest, hashed tokens and one-environment token scope. The distinction matters: encryption protects stored values, while token scope limits what a client can retrieve. Neither replaces careful handling of the master key.
The project explicitly does not auto-reencrypt values when the master key changes. That limitation is in the setup instructions and should be understood before relying on the system. It is a good example of why an architectural note should include constraints as well as features.
Inspect the envless README and source ↗ · See the project summary →
XIFF · Python / geometry
Describing a drawing by intent instead of path coordinates
SVG paths are precise, but a small drawing often starts with simpler instructions: draw a line of this length, bend a curve here, mirror a shape, repeat it around a point. XIFF is an experimental language that accepts that kind of description and writes a PNG. It is deliberately small and dependency-free at runtime. Its repository includes examples, tests and the renderer.
The geometry boundary
A stroke begins in a local x-axis coordinate system. A curve can be specified by length, bend position and bend amount; the engine solves the quadratic curve and transforms it into the canvas. Reusable groups, mirror, repeat and placement let a small set of instructions make a more complex drawing. That puts the coordinate work in the engine while keeping the input readable to a person editing a shape.
This is a tradeoff, not a replacement for SVG. A compact language is easier to compose for some shapes, but it exposes limitations when a drawing needs fine control. The XIFF README labels the tool an experimental alpha and shows logo studies that intentionally demonstrate where its present primitives are less concise. The examples make both the benefit and the boundary inspectable.
Read the XIFF source, examples and status ↗ · See the project summary →
From notes to evidence
These notes describe publicly inspectable projects. For a broader view of my work, use the project notes and GitHub profile. If you are evaluating an individual developer for a role or project, the direct-hire page offers a practical way to start.