Developer Defends Mir “Why Mir”
Canonical developer Alan Griffiths has been blogging a lot in recent days about the Mir server. He has been trying to get the community to support Mir and potentially even add native Wayland client support. His latest post is titled “Why Mir” with many still wondering why they should care about Mir when Wayland has proven to be the tested and widely adopted path.
Griffiths talks about Mir’s abstractions, Mir’s support for Mesa KMS/X11 and Android HWC graphics (although those are all possible with Wayland too), Vulkan support “beyond proof of concept” even though it is not public yet and Wayland supports Vulkan, and Mir as a display server providing “sensible defaults.”.
Because the work was funded with Canonical resources, features important to Ubuntu Phone and the Unity8 desktop have progressed faster and are more complete than others.
When Mir started, we needed a client-server communication mechanism (and Wayland was not in the state it is today). We made something that worked well enough (libmirclient) and, because it is only a small, deliberately isolated part of the whole, we could change it later. We never imagined what a “big deal” that decision would become.
Those who want to read more can visit the post Why Mir. But after extensively monitoring the Linux graphics landscape, about the only selling point I see is its sensible defaults making it easier for developers to get new shells up and running on Mir quickly. We have already seen that bringing up Wayland compositors can be widely done, though, and more work has been done (and is still ongoing) across the Wayland ecosystem to enable more code sharing. One of the major problems for Mir, however, is that XMir and Mesa driver support are not mainline and thus are not readily available outside Ubuntu packages.
Links
- http://www.phoronix.com/scan.php?page=news_item&px=Mir-Why-Mir
- http://voices.canonical.com/alan.griffiths/2017/04/12/why-mir/
Nine years later: what actually happened
The text above came a week after the funeral
The irony that makes sense of this post is only visible with a calendar in hand. On April 5, 2017, Mark Shuttleworth announced that Canonical would end Unity, convergence, and the Ubuntu Phone effort — and that Ubuntu 18.04 LTS would return to GNOME. The post “Why Mir”, by Alan Griffiths, dates from April 12.
In other words: the developer was defending the project’s reason for existing one week after the company abandoned the bet that justified it. This should not be read as stubbornness — it is the difference between a business decision and the technical conviction of the person who wrote the code.
Wayland won, and by a wide margin
The prediction implicit in the original text came true in full. Wayland is now the Linux desktop standard: GNOME has used it by default since Fedora 25, KDE Plasma 6 did the same, and major distributions have already removed or are removing the X11 session.
The problems the 2017 post pointed out in Mir — a driver not integrated into Mesa’s main tree, XMir outside mainline, availability limited to Ubuntu packages — were exactly the right ones. Without being upstream, a graphics server does not survive.
But Mir did not die — it became something else
And here is the part almost nobody followed. Instead of ending the project along with Unity8, Canonical repositioned it completely. Mir stopped being a competing graphics server for Wayland and became a library for building Wayland compositors.
It is a complete reversal of roles: the protocol the original post treated as a rival became the protocol Mir speaks. Today it is a set of libraries for anyone who wants to write a graphical shell without starting from scratch — with touch, mouse, and tablet input, multiple monitors, and client-server communication already handled.
Where it lives in 2026:
- Ubuntu Frame — Canonical’s kiosk and digital signage product, maintained by the same team as Mir. If you have seen a self-service kiosk running Linux, it may have been this.
- Embedded devices and IoT — the niche to which Canonical redirected its effort in 2017.
- Miriway — a Mir-based Wayland compositor for those who want to use it as a desktop.
- And a delightful detail: starting with version 2.26, Mir began rewriting its frontend for Wayland in Rust, work that continued in 2.27. The project that lost the desktop battle in 2017 is among those adopting the language the kernel only fully accepted in 2026.
Alan Griffiths, author of “Why Mir,” stayed with the project all these years.
The lesson that remained
Mir’s case is the clearest example of a free software rule: being technically better does not win; being integrated upstream does. Mir had real technical arguments — clean abstraction, sensible defaults, multiple back-end support. It lost because its driver was not in Mesa on the next person’s distribution, and because the company funding it changed its mind.
And there is a second, more optimistic lesson: good code is rarely wasted. Mir did not become an archived repository — it became infrastructure for something else, in a market where nobody expected to find it.
For the full context of this transition: Wayland will be the new X, from the same period, and Linux turns 35, which tells how the desktop got here.