HNHacker News
TopNewBestAskShowJobs

lherman

15 karma · joined November 10, 2020

submissionscomments
lherman··on Moving a Rust WebRTC SFU to thread-per-core: 70ms → 10ms P99.99 latency
PulseBeam is an open-source, lightweight WebRTC SFU server. Somewhere between LiveKit and mediasoup, written in Rust.
lherman··on Show HN: Lightweight, Opinionated WebRTC SFU in Rust
PulseBeam is an early-stage, open-source WebRTC SFU in Rust, built for simplicity. See https://pulsebeam.dev and https://github.com/pulseBeamDev/pulsebeam. Key features:

    Signaling: WHIP/WHEP superset over stateless HTTP, supporting custom low-frequency signaling.

    Platform Support: Requires only basic WebRTC compatibility, targeting a wide range of devices, including embedded systems.

    No TURN: Uses public host candidates with TLS (planned).

    Ports: Single UDP and TCP port (planned).

    Codecs: H.264 baseline (up to 4.1) and Opus for hardware acceleration.
Early project with a basic demo. Feedback and contributions welcome!
lherman··on Show HN: PulseBeam – Simplify WebRTC by Staying Serverless
That would be ideal. But, I don't believe there is a strong incentive for many existing network to be upgraded without using NAT.

In regards to TURN, it is used only as a fallback.

lherman··on Show HN: PulseBeam – Simplify WebRTC by Staying Serverless
Yes, self-host signaling server here: https://github.com/PulseBeamDev/pulsebeam-server-foss

LMK if you have any issues, contributions welcome

lherman··on Show HN: PulseBeam – Simplify WebRTC by Staying Serverless
We don't have an example of CRDT with PulseBeam yet. But, CRDT itself is just a data structure, so you can use PulseBeam to communicate the sync ops (full or delta) with a data channel. Then, you can either use y.js or other CRDT libraries to manage the merging.

Yes, the plan is to use JWT for both the client and server side.

OIDC is not on the roadmap yet. But, I've been tinkering on the side related to this. I think something like an OIDC mapper to PulseBeam JWT can work here.

I'm not envisioning integrating into a decentralization ecosystem at this point. The scope is to provide a reliable service for 1:1 and small groups for other developers to build on top with centralization. So, something like Analytics, global segmented signaling (allow close peers to connect with edge servers, but allow them to connect to remote servers as well), authentication, and more network topology support.

That's correct, if the client is reachable by a public IPv6 (meaning the other peer has to also have a way to talk to an IPv6), then STUN and TURN are not needed. ICE is still needed but only used lightly for checking and selecting the candidate pair connections.

lherman··on Show HN: PulseBeam – Simplify WebRTC by Staying Serverless
Yeah, if WebRTC is the only communication channel, TURN will be needed, especially doing p2p. LMK your experience if you have any questions, hit me up on discord @lherman_cs
lherman··on Show HN: PulseBeam – Simplify WebRTC by Staying Serverless
Awesome! Yeah I have been doing WebRTC for a while and it's still always a pain to get this stuff up and manage it, I would love to hear what you're building and have you give it a try, feel free to hit me up on discord, username @lherman_cs