The consensus among experts is that enterprise architecture is more than just a fad. If that is true, it seems to be at a crossroads. If and when it is completed, an operational federal enterprise architecture would represent a triumph of engineering and money.

Implementing architectures at the agency level is already no small feat, said Randy Hite, director of information technology architecture and systems issues at the Government Accountability Office. At the federal level, "the challenges have been increased by an order of magnitude," he said.

The federal enterprise architecture has struggled for resources and attention since its creation. For a long while in 2004, the effort looked rudderless, bereft of a chief architect in the Office of Management and Budget. During that dry time, the House tried to eliminate the position of chief architect.

Despite the setbacks and scarce resources, proponents say the federal enterprise architecture is on the cusp of something great. "It is time to move from the conceptual development and initial deployment phases," a CIO Council document states. The moment has arrived to take the federal enterprise architecture "into full implementation and operation states."

Attention has focused mainly on the five reference models, four of which have complete first iterations. But the reference models aren't an architecture. They're classification schemes — taxonomies — of performance measurements, business functions, applications and technical standards. A task force is revising the fifth model — the data reference model — to include an operational component for data exchange, but it is not yet complete.

Agencies use the models mainly to satisfy OMB's budget process requirements. They have few operational elements. But "the reference models are tools to be used in order to do things that ultimately result in an architecture," said Mike Tiemann, a senior associate at Booz Allen Hamilton and an architecture proponent.

But even as a tool for budget analysis, the reference models fall short. Conceived as parts of a whole, the five models were built one by one and have ended up disjointed — their connections with one another get lost. "That would have been great if they had done it in a fully integrated fashion, but it wasn't practicable," Tiemann said. What's missing is a way to link the models so definitions occur in a consistent, shareable and connectable way, he added.