Open Source Summit + ELC North America 2019

High performance internationalization support is coming to the main Linux filesystems. Historically, file and directory names have always been treated as opaque byte sequences by Unix kernels, in opposition to what was always done in Windows and MacOS. This design decision reduces the complexity of filesystem implementations, allowing them to defer meaning (encoding) to upper userspace layers, but it falls short, on the other hand, when the encoding information is required in kernelspace, for instance when case-insensitive filesystem lookups are needed. This talk will present the reasoning and use cases for in-kernel case-insensitive name lookups and the new Unicode interface for Ext4 that is being merged in the Mainline Linux kernel.