Tell HN: A Proposal to Modernize Xorg as a Protocol-Only Graphics Layer
5 points
11 hours ago
| 2 comments
| HN
The Linux desktop has spent more than a decade transitioning toward a new graphics stack. Wayland brings many advantages, especially for mobile-style security and simplicity. But in this process, we are quietly losing something valuable: the distributed, protocol‑driven, transport‑agnostic ideas that once made the Linux graphics model unique.

This is not nostalgia. These capabilities matter for remote work, automation, multi‑machine workflows, thin clients, cloud desktops, and future distributed systems. They are not “legacy features”; they are architectural strengths that may become important again.

The problem is not Wayland itself, but the fact that it was never designed to support these use cases. Its philosophy is intentionally local, single‑user, and compositor‑centric. That is perfectly valid for mobile devices, but it leaves a gap for desktop and distributed environments.

Xorg, on the other hand, suffers from an aging implementation, not an outdated philosophy. Its core ideas—protocol‑based rendering, remote execution, composability, and transport independence—remain relevant. What we lack is a modern, minimal, protocol‑only successor that preserves these strengths without carrying Xorg’s historical baggage.

Such a project would not need to replicate Xorg’s entire feature set. It would not need server‑side rendering, fonts, input methods, window management, or security policy. It would simply define a clean, modern protocol and a stable abstraction layer. Existing compositors could implement it. Existing drivers would not need to change. Mesa would not need major redesign. The engineering effort is far smaller than rewriting a full graphics stack.

This is not a call to replace Wayland. It is a call to acknowledge that the Linux desktop may need more than one graphics model. A protocol‑first, implementation‑agnostic layer could coexist with Wayland, complement it, and preserve capabilities that would otherwise disappear.

If no one starts this work, the industry will naturally converge on mobile‑style graphics architectures, and the distributed capabilities of the past may be forgotten for a long time. But if someone begins a modern protocol‑only successor to Xorg, the community may finally have a path that balances simplicity with the flexibility the desktop once had—and may need again.

wmf
8 hours ago
[-]
Phoenix: A modern X server written from scratch in Zig, 20 days ago, 445 comments https://news.ycombinator.com/item?id=46380075
reply
powerwordtree
52 minutes ago
[-]
Thanks for the link — Phoenix is a great project, and it clearly shows that a clean, modern X11 implementation still resonates with many developers.

What I’m talking about is something different. My post is not about building an implementation myself, but about advocating for a direction. The idea is not to keep Xorg’s implementation or its internal abstraction layer, but to preserve the design philosophy of X itself — the distributed, protocol‑based, client/server model that enables remote graphics, composability, and implementation diversity.

A key advantage of following the X philosophy is that we don’t need to change GPU drivers, because modern Linux already has Mesa/KMS/DRM as the real driver layer. A new protocol and a lightweight, strictly‑scoped abstraction layer could sit above that, combining X’s distributed design with modern techniques from Mesa, without inheriting X11’s legacy complexity. This layer would not implement any concrete rendering or window system logic — only provide the minimal interface needed for distributed graphics and multiple implementations to coexist.

This avoids the “Android-style” direction where the graphics stack becomes local-only and loses distributed capabilities. Instead, it keeps the door open for a healthier, multi‑implementation ecosystem.

(English is not my native language, so this reply is translated with AI.)

reply
bigyabai
11 hours ago
[-]
This is AI-generated nonsense. It makes 100x more sense to write a greenfield reimplimentation of the Xorg display server but you wouldn't know that asking an LLM to copy Wayland's design principles.
reply
powerwordtree
11 hours ago
[-]
Thanks for the comment. Just to clarify: the text was originally written in another language and I used an AI tool only to translate it because my English is not good enough for long technical writing. The ideas and arguments are my own, not generated by the model.

I’m not advocating copying Wayland or rejecting a greenfield implementation. My point was simply that a protocol‑first approach deserves to be part of the discussion, especially for use cases that Wayland intentionally doesn’t target.

reply
dang
1 hour ago
[-]
Ah, that's the problem. HN readers are super sensitive to any traces of AI-generated language in the comments. (Ironically, they often hallucinate it even when it isn't there—but that's another story.)

It turns out it's actually much better to just write in your own words and in your own voice, even if it's full of mistakes. We want to hear you, not a generated or filtered version of you.

Other explanations here in case helpful: https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que...

reply