Data sharing and interoperability were the inherent request of GIS. Three ways of data sharing that were data format conversion, data interoperability and direct data access had many shortcomings respectively. So we presented a new framework of data sharing and interoperability named Spatial Data Memory Engine (SDME) in this paper. It introduced different data of multi-formats, multi-systems and several spatial database engines into the memory container, created a memory database, and established a unified data structure which eliminated the differences of all spatial data. This made heterogeneous spatial data homogeneous, distributed spatial data localization, which enable different data to have the unified and transparent structure to all GIS users. We employed the memory dispatch policy of "layered vertically, grided horizontally" to resolve the contradiction between increasing spatial data and the limited memory, used incremental transmission mechanism to realize data synchronization, and gave three ways to handle the different user's conflicts. We also achieved file-based multi-users concurrent accessing by SDME. An example was given to illustrate the superiority of SDME at the end of this paper. In fact, SDME was a thorough unified framework of data sharing and raised the theory of spatial data sharing and interoperability to a new height.