Skip to content

ROADMAP.md

Latest commit

 

History

History
136 lines (94 loc) · 4.04 KB

ROADMAP.md

File metadata and controls

136 lines (94 loc) · 4.04 KB

RootStream Roadmap

This roadmap is not a contract and it is not the source of truth for current support.
Use docs/PRODUCT_CORE.md for the supported product definition, docs/SUPPORT_MATRIX.md for current surface maturity, and docs/microtasks.md for active execution work.

This document should describe future work, hardening targets, and deferred ambitions only.

Core principles (these NEVER change):

  • Linux-first
  • Low latency over convenience
  • Minimal dependencies
  • No custom cryptography
  • Clarity over cleverness

v1.1 – Linux Core Path Hardening

Goal: End-to-end Linux ↔ Linux streaming that is boringly reliable.

Features

  • Harden the existing Linux peer/client path

    • VA-API decode path
    • Basic frame presentation (SDL2 or DRM/KMS)
    • Input capture + injection (keyboard/mouse)
    • Clean connect/disconnect flow

  • Deterministic latency instrumentation

    • Host-side timestamps per stage (capture → encode → send)
    • Optional log mode: p50/p95/p99 latency over time (host loop)
    • Client-side timestamps per stage (recv → decode → present)
    • Debug overlay (FPS + latency) on client

  • Identity & pairing

    • Stable device identity key storage
    • Human-readable peer fingerprint (for tray/CLI)
    • Document backup/restore of identity keys

Docs

  • Update ARCHITECTURE.md with client-side architecture
  • Add basic troubleshooting for:
    • Client cannot decode
    • Black screen / no frames
    • Input not working

v1.2 – NVIDIA & Audio

Goal: Make RootStream the obvious choice for NVIDIA users and add solid low-latency audio.

NVIDIA path (NVENC/NVDEC)

  • Direct NVENC support on host

    • Use NVIDIA SDK directly (no VA-API shim)
    • Latency-focused presets
    • Fallback/disable on unsupported GPUs

  • NVDEC support on client (optional but ideal)

    • Detect NVIDIA hardware decode
    • Benchmark vs VA-API path

Audio streaming

  • ALSA-based audio capture on host
  • Opus encoding (small frame size: 2.5–5 ms)
  • Audio transport in protocol
  • Client-side Opus decode and playback
  • A/V synchronization (shared clock model)
  • Option to disable audio entirely

Docs

  • GPU-specific documentation (Intel/AMD/NVIDIA)
  • Audio troubleshooting (desync, stutter, missing audio)

v1.3 – Multi-monitor & polish

Goal: Handle real desktops, not just single-monitor lab setups.

  • Multi-monitor selection

    • Enumerate outputs/CRTCs
    • Choose which monitor to stream
    • Document behavior on hotplug / unplug

  • Better host UX

    • Tray UI for monitor selection
    • Quick toggle for “stream full desktop” vs “single monitor”

  • Stability & polish

    • Fuzz-ish testing of packet handling
    • Crash-resilient service mode

v2.0 – Protocol as a Platform

Goal: Turn RootStream into a clean reference implementation and protocol that others can target.

  • Formalize wire protocol in PROTOCOL.md

    • Message types (control, video, audio, input)
    • Version negotiation
    • Compatibility rules

  • Headless / automation mode

    • rootstream host --headless --card /dev/dri/cardX
    • Use cases: render nodes, CI visualizations, remote GPU workloads

  • Harden the existing Windows client path into a supportable client-only lane

    • Decode + input only
    • LAN/VPN use, no cloud

  • Experimental hooks (documented)

    • Pre-encode/post-decode hooks
    • Input filters (for power users only)

Out-of-Scope (for now)

These are explicitly not priorities, to protect RootStream’s identity:

  • User accounts, central servers, or any “cloud” control-plane
  • In-browser/WebRTC client
  • “Install-everything for everyone” GUI installer
  • Generic remote desktop for every OS under the sun

RootStream focuses on:

Secure, low-latency, Linux-centric game streaming
with transparent architecture and minimal moving parts.