All physical implementations of quantum bits (qubits), carrying the information and computation in a putative quantum computer, have to meet the conflicting requirements of environmental decoupling while remaining manipulable through designed external signals. Proposals based on quantum optics naturally emphasize the aspect of optimal isolation [1][2][3], while those following the solid state route exploit the variability and scalability of modern nanoscale fabrication techniques [4][5][6][7][8]. Recently, various designs using superconducting structures have been successfully tested for quantum coherent operation [9][10][11], however, the ultimate goal of reaching coherent evolution over thousands of elementary operations remains a formidable task. Protecting qubits from decoherence by exploiting topological stability, a qualitatively new proposal due to Kitaev [12], holds the promise for long decoherence times, but its practical physical implementation has remained unclear so far. Here, we show how strongly correlated systems developing an isolated two-fold degenerate quantum dimer liquid groundstate can be used in the construction of topologically stable qubits and discuss their implementation using Josephson junction arrays.Any quantum computer has to incorporate some fault tolerance as we cannot hope to eliminate all the various sources of decoherence. Amazing progress has been made in the development of quantum error correction schemes [13] which are based on redundant multi-qubit encoding of the quantum data combined with error detection-and recovery steps through appropriate manipulation of the data. Error correction schemes are generic (and hence are applicable to any hardware implementation), but require repeated active interference with the computer during run-time; the delocalization of the data, often in a hierarchical structure, boosts the system size by a factor 10 2 to 10 3 . Delocalization of the quantum information is also at the heart of topological quantum computing [12], however, the stabilization against decoherence is entirely deferred to the hardware level (hence it is tied to the specific implementation) and is achieved passively. In searching for a physical implementation of topological qubits one strives for an extended (many body) quantum system where the Hilbert space of quantum states decomposes into mutually orthogonal sectors, each sector remaining isolated under the action of local perturbations. Choosing the two qubit states from groundstates in different sectors protects these states from unwanted mixing through noise; protection from leakage within the sector has to be secured through a gapped excitation spectrum. As no local operator can interfere with these states, global operators must be found (and implemented) allowing for the manipulation of the qubit state.A promising candidate fulfilling the above requirements is the quantum dimer system [14][15][16]: recent quantum Monte Carlo simulations of the dimer model on a triangular lattice provide evidence for a gapped liquid ...