<p>The box-counting method is widely applied to quantify the fractal dimension of fracture networks, a procedure that, while seemingly straightforward, presents several challenges. Over the past four decades, conflicting conclusions have emerged regarding whether fracture networks exhibit genuine fractal properties. Some researchers argue that a fracture network is fractal, while others maintain the opposite. This study critically re-examines these debates by addressing five central issues: the validity of the classical definition of fractal dimensions, the permissible range of values, the role of self-similarity, the constancy of the fractal dimension across scales, and the influence of multiple fracturing episodes. Analytical derivations show that applying the classical definition invariably yields a dimension of unity, indicating limited practical relevance. Simulations with synthetic discrete fracture networks demonstrate that box-counting curves are often nonlinear, with local slopes occasionally producing results difficult to interpret physically. Residual and local slope analyses further highlight the influence of grid resolution, sampling density, and fracture organization on the stability of the estimates stability. Multi-scale case studies, from seismic to thin-section imagery, corroborate these findings, revealing that apparent linearity can obscure underlying complexity. Building on these insights, a refined interpretive framework is proposed to assess fractality and estimate fractal dimensions, helping reconcile previous inconsistencies and offering methodological guidance for applications in reservoir characterization and structural geology.</p>