Design criteria for meso- and cloud-scale atmospheric models have typically centered around (i) accuracy for short-term integrations (hours to days), (ii) efficiency, and (iii) robustness - especially for weather prediction applications.  While these criteria are being used in the WRF model design, we are also attempting to address the needs of the atmospheric chemistry and air quality communities with the WRF model.  The primary criteria introduced by these needs is mass conservation and conservation of transported species in the continuous and discrete dynamical equations.  Mesoscale and cloudscale models are necessarily nonhydrostatic, and the need for efficiency and robustness have, in the past, led to model equations that did not conserve mass or scalar quantities.  Within the WRF model development effort at NCAR, we have constructed new integration techniques that conserve mass and scalar quantities while preserving the needed efficiency and robustness of existing formulations.  I will briefly outline both the old (non-mass conserving) and new (mass conserving) formulations.  There are several mass-conserving formulations being considered, including one based on semi-Lagrangian integration techniques.  I will briefly discuss the decisions that are still to be made concerning equations and integration techniques for the WRF model, our general mesoscale and cloudscale model design philosophy, and the advantages/disadvantages of online versus offline chemical/air-quality modeling approaches.