Global Navigation Satellite Systems (GNSS) provide ubiquitous, continuous, and reliable positioning, navigation, and timing (PNT) information and are a core component of modern infrastructure. However, a revisit of GNSS constellation design decisions —based on the precursor Global Positioning System (GPS) design conceived in the 70s— is justified by recent developments: (1) Longer satellite lifetimes, (2) Demand for stronger signals, (3) Increased risk of space debris, (4) Advancements in satellite production lines and components enabled by the New Space economy, (5) 10-fold reduction in launch costs. Our analysis uses up-to-date information on satellite buses, end-of-life practices, navigation payload power consumption, and launch costs to analyze design trade-offs as signal power levels increase.The recent interest in lunar exploration has also highlighted the need for the development of Lunar based navigation infrastructure to overcome the limitations of weak GNSS signal tracking and ground tracking networks. Tradespace exploration and a multi-objective evolutionary algorithm are used to explore the design space of an Earth and a Lunar GNSS respectively. The results are mapped to design decisions by sensitivity analysis and association rule mining. The position and clock bias performance achievable in the vicinity of the Moon with a Lunar GNSS is simulated by Linear Covariance (LINCOV) techniques, under a variety of assumptions of orbit determination error, measurement types, and user dynamics and clock quality. This analysis is conducted at different stages of constellation deployment using Epoch-Era analysis. The Earth GNSS study finds that constellations at an orbit altitude of ~2 Earth radii can outperform existing GNSS in terms of cost, robustness, and User Navigation Error (UNE). Great candidates for a Lunar GNSS are Walker constellations in near-circular polar orbits at an altitude of ~2 lunar radii. It is shown that the absolute positioning goal of 40cm (as defined by the International Space Exploration Coordination Group (ISECG)) can be achieved by a sub-constellation of at least 16 satellites if pseudorange and delta range measurements are used and if the orbit determination error sigma is one meter or better. These insights should be considered when designing future generations of GNSS.