PeerJS vs Discord

Every entry in PeerJS's feed is a release candidate, and each one repeats the bug list of the one before it — the unreliable close event and the data channel ordering fix appear in all four, spanning fourteen months. The genuinely new content per release is thin: a CSP fix removing the need for unsafe-eval, Blob support inside nested objects, and dependency updates. The most recent publication is from April 2024.

The changelog reads as a project maintained rather than developed, and one that never quite finishes a release cycle — candidates accumulate, notes carry forward, and stable releases do not appear in this feed. For a WebRTC wrapper this is a specific kind of risk: the browser layer underneath moves regardless, and the fixes that do land here are reactions to it, such as dropping unsafe-eval so the library survives strict Content Security Policies.

There is no signal in these entries about what comes next, and no publication in roughly two years; the honest read is that the release cadence has stopped rather than that a particular change is pending.