As more scientist turn to HPC, new storage technologies enter the market, and lines blur between HPC and cloud computing, it’s getting harder and harder to find “the one true storage system” that does it all. Specialized storage systems offer the promise of higher efficiency and scientific productivity, but they also threaten key tenets of portability, interoperability, and data stewardship. Is it possible to realize the dream of specialized storage systems without sacrificing long-standing best practice for HPC data management? Can we appropriate ideas from other fields to help bridge this gap?