LoginSign Up
← Back to all sparks
R

Red5

MEETINGS
Velocity2.5

Open-source media server for live streaming and video

www.red5.net ↗

Red5 ships constantly and explains almost none of it.

media-serverrtmpav1streaming-performancejavaundocumented-releases
◆Current state
Red5 is a Java media server built around RTMP ingest and delivery. Releases come every few days — nine tags between early May and early July — but most publish nothing beyond an auto-generated link to the commit range. Two version schemes run side by side, with v2.0.16.46 appearing in July while the v2.0.3x series was still incrementing through June, so the tag number alone does not tell you which line a build belongs to.
◆Where it's heading
Where changelogs do exist, the work is concentrated in the RTMP hot path: cutting CPU waste and allocation churn on ingest, then addressing thread oversubscription three releases later. The one capability addition visible in this window is AV1 support. Everything else that is described is dependency maintenance — Spring, Bouncy Castle — and build plumbing like on-demand Docker image pushes. The pattern is a small maintainer group optimising a mature server rather than repositioning it.
◆Prediction
Expect the RTMP performance thread to continue and AV1 handling to be refined in follow-up releases. The undocumented releases make anything beyond that guesswork — the changelogs would need to carry content before a direction could be read from them.

◆Recent moves

  1. 24d ago

    Red5 2.0.16.46 ships with no published changes

    Tagged with nothing but a link to the commit range against 2.0.16.44. It also belongs to the 2.0.16.x scheme rather than the 2.0.3x/2.0.40 line that was releasing through June, which makes the relationship between the two trains impossible to read from the release notes.

    View source ↗
  2. 1mo ago

    Red5 2.0.40 ships with no published changes

    An empty release note pointing only at the diff since 2.0.39. Coming five days after the AV1 release, it is most likely follow-up work, but nothing in the entry confirms that.

    View source ↗
  3. 1mo ago

    Red5 2.0.39 adds AV1 support

    The only capability addition described in this window: AV1 lands in a server whose recent history is otherwise RTMP performance work. Codec coverage is table stakes for a media server rather than a repositioning, but it is the first change here that adds something rather than making existing paths cheaper.

    View source ↗
  4. 2mo ago

    Red5 2.0.38 fixes RTMP thread oversubscription

    A performance change addressing thread oversubscription in RTMP handling, two releases after the ingest hot-path work. Together they suggest a maintainer working through the concurrency model of the ingest layer methodically rather than reacting to individual reports.

    View source ↗
  5. 2mo ago

    Red5 2.0.37 ships with no published changes

    Released three hours after 2.0.36 with only a commit-range link. The timing points at a fix for something the previous release introduced, but the notes say nothing.

    View source ↗
  6. 2mo ago

    Red5 2.0.36 cuts CPU waste in RTMP ingest

    The start of the current performance push, targeting hot-path CPU waste and allocation churn on RTMP ingest. For a server whose cost profile is dominated by concurrent inbound streams, allocation churn on that path is the right place to look.

    View source ↗