cid 3d63abbda768675443eb5121840e182c5fb35f23639f95a725741ef7e6a7491a · original · open access · via openalex
Title: Replication package: Female Participation in the Linux Kernel Authors: Anonymous Authors Venue: Zenodo (CERN European Organization for Nuclear Research) Year: 2026 Canonical: doi:10.5281/zenodo.23828638 URL: https://doi.org/10.5281/zenodo.23828638 Collector: openalex OA: true ABSTRACT Replication package of the paper Female Participation in the Linux Kernel: A Perspective from Mailing Lists and Patches. The package rebuilds every figure, table and number of the paper from the 350 Linux kernel mailing lists of the LKML5Ws dataset (29.4 million e-mails, 1996–2026). The pipeline: Identity consolidation: merges 415,582 sender identities into 217,175 persons, using all 822 historical versions of the kernel .mailmap, Gmane address de-obfuscation and name normalization. Gender inference: labels each person's name with two local LLMs (Llama 3.2 3B and DeepSeek-R1-Distill-Qwen-7B, through llama.cpp), which may abstain. gender-guesser is used as a baseline, and a stratified sample is compared with NamSor. Participation analysis: share of women per year, among newcomers and per Git trailer (Signed-off-by, Reviewed-by, Acked-by, Tested-by), and retention. Homophily: observed/expected (O/E) ratios for reply, thread and trailer ties, overall, stratified by list and over time. Pseudo-anonymization: builds the released dataset, whose hashes are compatible with the anonymized LKML5Ws release. The non-anonymized mailing lists are not included. They are produced with the LKML5Ws replication package (https://doi.org/10.5281/zenodo.20434788), following the manual step described in README.md. The resulting dataset of 141,171 active participants with gender labels is published as a separate Zenodo record. Contents: Python code (src/, paper/, future_work/), the notebook generator (tools/), the script that rebuilds the kernel .mailmap history, and a README.md that maps each result of the paper to the script that produces it.