The need for communication is highest in disaster scenarios when the infrastructure is also adversely affected. A recent protocol for ad hoc communication, the BATMAN protocol, is dependent on minimal infrastructure, in the form of mesh nodes that are used as access points, or nodes acting as an intermediary in a multi-hop connection. While BATMAN works well in a scenario in which there is a multihop path from senders to receivers at all times, it will drop the packets in intermittently-connected networks. Moreover, although implementation on a device is essential as a proof of concept, performing large scale evaluations requires a simulation platform in which variations in the operating environment can be studied. This paper is about adding the store-and-forward mechanism to the routing component in BATMAN nodes, to overcome intermittent connectivity through mobility. We describe an extension of the protocol, SF-BATMAN, that has been implemented in an interoperable manner with BATMAN, i.e. with no added signaling, and no change of basic BATMAN settings. We have implemented SF-BATMAN in a packet level simulator (NS3), and demonstrated its performance in a scenario that consists of two regions of connectivity: a well-connected mesh network and a set of sparser subnetworks. We show that the added capability enhances the performance of BATMAN, through an increase of the delivery ratio by 20% with a lower overhead, while it exhibits a similar latency in comparable network scenarios.