Quantifying dissimilarities between pairs of networks is a challenging and, at times, ill-posed problem. Nevertheless, we often need to compare the structural or functional differences between complex systems across a range of disciplines, from biology to sociology. These techniques form a family of graph distance measures, and over the last few decades, the number and sophistication of these techniques have increased drastically. In this talk, I offer a framework for categorizing and benchmarking graph distance measures in general: the within-ensemble graph distance (WEGD), a measure that leverages known properties of random graphs to evaluate the effectiveness of a given distance measure. In doing so, I hope to offer an invitation for the development of new graph distances, which have the potential to be more informative and more efficient than the tools we have today. I close by offering a roadmap for identifying and addressing open problems in the world of graph distance measures, with applications in neuroscience, material design, and infrastructure networks.