In this paper we compare the performance of tree different dynamic traffic engineering algorithms exploiting path diversity in the Internet, TEXCP, TRUMP and MIRTO. We passed through a thorough implementation phase of these algorithms solving a number of issues related to protocol implementation that allows a complete analysis in real traffic settings in real networks. We discuss the performance of such protocols and some difficulties encountered during their implementation. Supported by an already mature standardization framework we conclude that dynamic yet stable traffic engineering is not only feasible but expected with rising interest by network operators.