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.)
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.
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.
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.
Phoenix: A modern X server written from scratch in Zig, 20 days ago, 445 comments https://news.ycombinator.com/item?id=46380075
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.)
the distributed, protocol‑based, client/server model that enables remote graphics, composability, and implementation diversity
This mostly sounds like Wayland to me. Anyway, build it.
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.
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.
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...