Due to the large amount of information (messages, articles, videos, music, images, etc.) generated and shared on social networking sites, users find themselves overwhelmed by infor- mation generated chronologically in their news feed. In addition, most of information may be irrelevant. Sorting social updates, in order of relevance, is proposed as a solution to help users quickly view and interact with information that may interest them. In this work, we study existing approaches in the area of sorting social updates and expose their limits and some open issues according to several axes: factors influencing information's relevance, prediction models of information's relevance, training and evaluation of prediction models, etc.