Online discussions are valuable resources to study user behaviour on a\ndiverse set of topics. Unlike previous studies which model a discussion in a\nstatic manner, in the present study, we model it as a time-varying process and\nsolve two inter-related problems -- predict which user groups will get engaged\nwith an ongoing discussion, and forecast the growth rate of a discussion in\nterms of the number of comments. We propose RGNet (Relativistic Gravitational\nNerwork), a novel algorithm that uses Einstein Field Equations of gravity to\nmodel online discussions as `cloud of dust' hovering over a user spacetime\nmanifold, attracting users of different groups at different rates over time. We\nalso propose GUVec, a global user embedding method for an online discussion,\nwhich is used by RGNet to predict temporal user engagement. RGNet leverages\ndifferent textual and network-based features to learn the dust distribution for\ndiscussions.\n We employ four baselines -- first two using LSTM architecture, third one\nusing Newtonian model of gravity, and fourth one using a logistic regression\nadopted from a previous work on engagement prediction. Experiments on Reddit\ndataset show that RGNet achieves 0.72 Micro F1 score and 6.01% average error\nfor temporal engagement prediction of user groups and growth rate forecasting,\nrespectively, outperforming all the baselines significantly. We further employ\nRGNet to predict non-temporal engagement -- whether users will comment to a\ngiven post or not. RGNet achieves 0.62 AUC for this task, outperforming\nexisting baseline by 8.77% AUC.\n