In this paper we report an experimental comparison between two Ad Hoc routing protocols on real networks. Specifically, we evaluate performance of OLSR and AODV either in indoor and outdoor environments on networks of 2-4 hops size with up to 8 nodes, representing realistic scenarios of few people exploiting the Ad Hoc network to share documents. In fact, as pointed out in [13], with current technology, benefits of Ad Hoc network will vanish beyond the Ad Hoc horizon of 2-3 hops and 10-20 nodes. Our analysis shows that with semi-static topology the proactive approach performs much better than the reactive from the efficiency and QoS standpoint, and it introduces a limited overhead. On the other hand, even in this simple scenario, AODV performances are often poor introducing delays of seconds in order to ping a node few hops away. To avoid possible proactive-protocol scalability problems, we also discuss our ongoing work to develop a proactive protocol with partial dissemination.