It pushes the incentives around as well, though. Unless there's some sort of marginal cost for each line of code they choose to keep proprietary, they'll inevitably do "the least work necessary" when they realize that some project is going to have components fitting these clauses, by by boxing up the entire effort under one of these clauses.
Ideally, with some sort of force opposing propriety, the incentive would be such that the government would think it in its best interests to granularize their code into small libraries, programs, and network services, so that they could open-source as much as possible, leaving tiny "classified kernels" as wrappers for libraries or plugins for apps.
For an idea of such a force: maybe create a "proriety credit" that a government department would to buy for each 1000 lines of code they refuse to release, whose funding goes to this effort to open-source everything else?
I was using LOC as a hand-wave-y abstraction; substitute something like "total AST-node information-theoretic entropy measured via compressed size of language-canonicalized source" if you want to be picky.
Ideally, with some sort of force opposing propriety, the incentive would be such that the government would think it in its best interests to granularize their code into small libraries, programs, and network services, so that they could open-source as much as possible, leaving tiny "classified kernels" as wrappers for libraries or plugins for apps.
For an idea of such a force: maybe create a "proriety credit" that a government department would to buy for each 1000 lines of code they refuse to release, whose funding goes to this effort to open-source everything else?