Skip to content

Engine/front split: a shared native core with separate fronts? #66

Closed

Description

@gnacho

I run Jellyfin at home and my main set is a rooted LG on webOS 5.5, so I have been watching this project with some envy. Every Jellyfin client on webOS today is the web app in a box, and none of them will ever feel like PlxNative does on an old panel.

Reading the source, there are two very different projects living in one repo. One is the webOS engine: starfish.c with the video plane, dynlib.rs with the SONAME candidate dance, the player, ABR, the FFmpeg glue, HLS. That part is knowledge that exists nowhere else, since LG's player APIs are not in OSE. The other is the Plex front: plex/, pms.rs, the QR sign-in, and a UI that speaks Plex types end to end.

My question is about the seam between them. Would you be open, at some point, to extracting the engine so separate fronts can sit on top of it? Your Plex front stays as it is, first and canonical. Everything past it is upside: a Jellyfin front, maybe an Emby one, whatever people end up bringing. The Jellyfin crowd has never had a native client on these sets, and some of them would show up as users and contributors if that changed. I may end up building one of those fronts myself, but the case does not rest on that.

The shape I would argue for is the full split: the engine as its own versioned core, fronts as separate projects on top. I will not propose the half-way version, a trait under the UI while Plex types keep flowing beneath it; half measures tend to make the final split harder, not easier. And the pace of the last few months suggests the cut is within what this repo already does: the search rework alone touched more surface than an engine extraction would. If I ever experiment in this direction, I would much rather do it on top of your tree than beside it, so asking first seemed the right move.

Activity

  1. GLinnik21 commented on Sep 3, 2026

    @GLinnik21
    Collaborator

    Thanks for the suggestion! I’ve actually thought about reuse too, but from a slightly different direction.

    My main goal with PlxNative has always been the user experience rather than building a particularly advanced playback engine. The playback/TV layer is actually the part I consider the most reliable already, and if the goal were a truly reusable media engine, mature projects like Kodi have years of work and testing across a huge range of devices behind them.

    What I had in mind instead was eventually making the native UI reusable across different backends — Plex, Jellyfin, Emby, etc. I think that’s where PlxNative could provide more value.

    For now, though, I’d like to get the Plex experience properly finished first, since that’s what I personally use and why I started the project.

  2. gnacho commented on Sep 4, 2026

    @gnacho
    Author

    Thanks, that makes sense, and I think you are right about the direction.

    The Kodi point is fair: a portable playback engine is a solved problem, and the thing PlxNative has that nothing else does is the native UI. Reusing that across media servers is a better trade than splitting the engine out, and it is the version that would actually help people on older sets, whatever they run on the server side.

    One thing I would keep an eye on when you get to that seam is the data model. The UI currently speaks Plex types end to end, and an abstraction written after the fact around Plex alone tends to stay Plex-shaped underneath. If a second backend is a design requirement from day one, even if only on paper, the boundary ends up neutral. I would be glad to help define that boundary, and I
    may well take the Jellyfin side when the time comes.

    Finishing the Plex experience first sounds right. The seam only matters once the app itself is where you want it.

  3. GLinnik21 commented on Sep 4, 2026

    @GLinnik21
    Collaborator

    Yeah, good point about the data model. I’ll keep that in mind while finishing the Plex side — I definitely don’t want to make it unnecessarily Plex-shaped if the UI can eventually be reused.

    And if you end up taking the Jellyfin side when we get there, that’d be awesome :)

  4. locked and limited conversation to collaborators on Sep 4, 2026
  5. converted this issue into a discussion #67 on Sep 4, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementPlanned improvement (ideas start in Discussions)

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions