Government attempts to avoid pain and vendor lock-in upfront will fail in the most drawn-out ways possible. For any interesting class of problem (i.e. non-trivial innovation) it isn’t possible to deductively design a reference architecture. Instead you must build and let the architecture emerge. You maximize the chance of getting it right by having multiple competing companies and programs with interoperability requirements at inception. Government Reference Architectures are the 21st century equivalent of Robert McNamara’s notorious Total Package Procurement (TPP), which produced failures like the F-111 and C-5A. TPP fell into the trap of trying to eliminate uncertainty and predict the chaos of the universe by inflexibly defining every program requirement and dollar spent, from R&D through production, before any work had actually been done. Chaos won.
Bankwatch 