TCP is not necessarily suitable for all contexts. It is well-known that, in some context, the "simple" fact of replacing TCP with another Transport protocol clearly improves the performance of the application, especially in terms of throughput. However, TCP remains the most widely used Transport layer (L4) protocol on the Internet: nearly 90% of applications are based on it. Why? Our analysis is that the need to rewrite the TCP application before using any new protocol "partly" explains the low adoption of new protocols despite TCP's performance limits. Indeed, constant modification of the application is tedious and, above all, a potential source of instability. This raises a second major question: can we, transparently to the application, replace at runtime TCP with another protocol X more suitable to the application requirements and the actual network condition? In this paper, we design and implement VTL to provide a concrete and practical answer to this question, i.e., the TCP reconfiguration at the runtime and the choice of the best alternative to it. Based on a series of experimental scenarios, we show the correctness of VTL and that it significantly improves TCP applications' performances.