The Fedora Council has rescinded its earlier decision to create Fedora AI Developer Desktop, an official edition of the distribution for developers using AI tools. Initially, all six members of the Council voted in favor of the project's creation, but after hearing criticism during a community discussion, two members changed their votes a few days later and voted against it. Since unanimity was not reached, approval of the decision has been postponed. The issue is planned to be resolved before the Flock 2026 conference, which will be held from June 14 to 16.
The dissatisfaction stems from the intention to use third-party kernel modules and proprietary components to support NVIDIA CUDA technology in Fedora AI Developer Desktop. The provision of third-party modules violates the project's established consensus regarding the third-party kernel module provisioning policy, while the provision of CUDA runs counter to the project's ideology, which discourages the promotion of proprietary software and the tying of solutions to a single vendor. Furthermore, the new release proposed provisioning LTS kernel branches, but it is unclear who will maintain them separately from the standard, regularly updated kernel packages maintained in Fedora.
According to the resigning board member, the change in Fedora's kernel module delivery strategy requires additional approval and expert opinions from engineers and lawyers. It is anticipated that the issue could be resolved by switching to the native Nova driver instead of NVIDIA's third-party open source modules, which are expected to be ready by the end of the year.
Initially, it was assumed that community members would contact the Governing Council representatives if they had concerns about upcoming decisions, but in reality, the Council was not informed of potential issues and voted based on information provided by the initiative's author. To prevent future Council decisions that run counter to community opinion, Jef Spaleta, the Fedora project leader, proposed publishing consensus Council positions before voting, allowing the community time to respond and attempt to change the Council's mind. Another option discussed is introducing a pre-vote during the discussion phase of Council issues, with the results published outside of the meeting minutes to allow the community to understand the Council's position before the final vote.
The Fedora AI Developer Desktop project aims to create additional atomically updated desktop editions of Fedora. Linux For developers using and developing AI technologies. The plan was to use the foundation of the Silverblue (GNOME) and Kinoite (KDE) distributions, supplemented with kernel modules, toolkits, platforms, and libraries for deploying AI models on a local system, as well as developing AI-based applications. A prototype build of Fedora AI Developer Desktop, based on Fedora Silverblue, is available for testing.
The ultimate goal is to provide an out-of-the-box AI solution that requires no manual configuration or additional software installation. Many AI frameworks typically require manual configuration and alignment of kernel versions, NVIDIA drivers, CUDA toolchain, and container runtimes to ensure proper operation. Fedora AI Developer Desktop provides a pre-tested working environment that allows you to immediately launch ready-made containers with pre-configured AI platforms and utilize GPU-accelerated AI model execution.
All AI tools and platforms will be delivered by default without sending telemetry and with cloud services disabled. To enable acceleration on NVIDIA GPUs, the package includes open-source kernel modules used in NVIDIA's proprietary drivers (NVIDIA plans to replace these kernel modules with the Nova driver in the future). CUDA Runtime or CUDA Toolkit can be installed on top of these open-source modules. In addition to NVIDIA GPUs, the plans also include support for systems with ARM, AMD, and Intel GPUs. Among the components included in the basic package are the pre-configured Goose AI agent and the Podman Desktop graphical interface for creating, running, and managing containers.
Source: opennet.ru
