What this stack does not do, and what is not yet proven:
| Area | Status |
|---|---|
| Several machines sharing one database | not supported: SQLite is one file on one disk. Use one machine, or a database server with another library |
| Many processes on one machine | supported (WAL, write locks, live changes across processes); a load test at 1, 4 and 16 processes is in progress |
| React component libraries | not compatible: TradJS has its own small JSX renderer |
<feDropShadow> in server-rendered SVG |
the HTML parser predates this element, so the server's version parses as fedropshadow; it works when the browser renders it |
| measure-fn's cost per call | not yet measured cleanly; the next version (6.1) is released only after it is |
The starter app (tradjs init) |
passes 30 steps in a real browser; in review for the next TradJS release |
| Production build check | 5 of its checks pass (hashed assets, no server code in browser bundles, byte-identical rebuilds, start from any folder); the rest are still to run |
| In progress: |
- live updates over websockets in the starter app, instead of polling;
- a real-browser check of development mode (edits, error pages, recovery);
- a tool that runs every code example in the sqlite-zod-orm README;
- the multi-process load test. Next: chapter 10, start an app.