The Lua community needs to learn to move on - hisham.hm

A recent upstream change removed archived tags for legacy LuaJIT releases, a move the author describes as "nuclear" but understandable given the flood of bug reports from decade-old code. This action directly cuts off the convenience of pulling old binaries from the official repo, pushing downstream users toward newer releases. For engineers, the immediate impact is a loss of a simple download path for legacy versions, requiring either self-hosting or abandoning them.

Historically, Lua developers have kept a wide swath of interpreter versions alive, from Lua 5.1 up through a tentative 5.5 and multiple LuaJIT branches. This practice arose after the Lua/LuaJIT split, where targeting Lua 5.1 served as a common denominator for both ecosystems. The consequence is that library maintainers must run test suites against a large matrix of versions, inflating CI time and maintenance effort.

The author argues for a shift toward an end-of-life (EOL) model similar to other dynamic languages, retiring old interpreter releases and encouraging library authors to do the same. Tools like Teal, which can emit Lua code for any target version, are presented as a way to bridge the gap without maintaining separate codebases. Engineers adopting this approach would need to integrate Teal into their build process or otherwise refactor code to target only current interpreter releases.

Adopting the EOL stance entails updating existing code to run on the latest PUC-Rio Lua and LuaJIT releases, which may involve API changes, dependency upgrades, or rewriting C bindings. The cost is primarily developer time spent on migration and testing against a reduced version set, but the payoff is a leaner maintenance surface and fewer legacy bug reports. Projects that cannot afford this migration, such as embedded systems locked to a specific Lua version, will lose official support and may need to maintain private forks.

The proposed change stops working for any software that deliberately embeds an old Lua interpreter and relies on its specific behavior, as those versions will no longer receive updates or be easily obtainable. In such cases, the engineering team must either accept the risk of unpatched bugs or allocate resources to maintain a private copy of the interpreter. For the broader ecosystem, the shift promises a more sustainable development model once legacy anchors are finally dropped.