Why AI-Generated Patches Drove Linux to Drop the 40‑Year‑Old AppleTalk Protocol

In the open‑source world, software is usually retired because it is either unused or unmaintained. Recently, the Linux kernel faced a rare situation: the removal of AppleTalk, a classic networking protocol that had been part of the kernel for four decades.

Jakub Kicinski, a Linux network subsystem maintainer, submitted a change that deleted nearly 4,000 lines of AppleTalk‑related code. The removal was motivated by an influx of AI‑generated fix patches for AppleTalk that received virtually no human review. As Kicinski explained, “Recently we received a wave of AI‑generated fix patches for AppleTalk, but no one reviewed them.”

AppleTalk, introduced by Apple in 1985, offered zero‑configuration networking that automatically discovered printers, file servers, and other devices on a local network. Its design foreshadowed modern discovery mechanisms such as Bonjour, mDNS, and Zeroconf. However, with the rise of TCP/IP as the universal standard, Apple gradually phased out AppleTalk, finally removing support from macOS X 10.6 Snow Leopard in 2009.

Even after Apple’s abandonment, the Linux kernel retained AppleTalk support through projects like Netatalk to serve legacy Macintosh hardware used by enterprises, educational institutions, and hobbyists. The code remained stable and cheap to maintain, allowing it to survive 17 years beyond Apple’s own support cycle.

The emergence of large language models (e.g., ChatGPT, Claude) capable of generating code has changed this balance. Developers increasingly rely on AI‑generated patches, submitting them directly to open‑source projects. For an obsolete module like AppleTalk, AI began producing numerous corner‑case fixes that accumulated without verification, testing, or dedicated maintainers. This created a maintenance burden that outweighed the protocol’s dwindling user base.

Faced with the rising cost of reviewing and integrating these AI patches, the Linux maintainers concluded that, since no one was using or maintaining AppleTalk, the pragmatic solution was to delete the code. The removal does not mean the protocol disappears entirely; the code will be migrated to an independent repository where classic Macintosh enthusiasts can continue its development, and a user‑space API (uAPI) will remain to avoid abrupt disruption for existing users.

AppleTalk is not the only legacy component being pruned. Recent development cycles have also targeted long‑neglected network protocols and drivers such as ARCnet, ISDN, AX.25, and various ham‑radio modules. While these technologies once served important roles, they are now being retired as industry standards evolve and user demand fades.

Looking ahead, the article warns that AI‑generated patches may increasingly target other “old‑fashioned” modules, raising the question: which legacy component will be the next to be overwhelmed and removed?