📝 SummarXiv

Abstract

We investigate the evolution rules and degree distribution properties of the Software Heritage dataset, a large-scale growing network linking software releases and revisions from open-source communities. The network spans over 40 years and includes about 6 billion nodes and edges. Our analysis relies on natural temporal and topological partitions of nodes and edges. A derived temporalized graph reveals a bow-tie-like structure and enables study of edge dynamics -- creation, inheritance, and aging -- together with comparisons to minimal models. In- and out-degree distributions and edge timestamp histograms expose regime shifts linked to changes in developer practices, notably in the average number of edges per new node. Without presupposing its validity, we estimate the scaling exponent under the scale-free hypothesis. Results highlight the sensitivity of a widely used estimation method to regime changes and outliers, while showing that partitioning improves regularity and helps disentangle these effects. We extend the analysis to the APS citation network, which also exhibits a major regime shift around 1985, though driven by distinct factors. Both cases illustrate how structural and dynamical transitions complicate conclusions about the existence and observability of a scale-free regime. These findings underscore the need for refined tools to study transient growth phases and to enable robust comparisons between empirical growing networks and minimal models.