[zbyszek]

Repetition isn't necessarily a bad thing. Whether compatible results can be obtained with independent analyses of the same data is surely an important indicator of success.

Of course that should not preclude the possibility of replication as described here.

I find the idea of threats to job security misguided. In the particle physics world some significant code (not everything, granted) has been made available for some time; e.g. Monte-Carlo simulators like Geant or lattice gauge theory codes like those from MILC, SciDAC or FermiQCD. Even highly optimised code has been available. Users are requested to cite the authors. No careers have been harmed, as far as I can tell.

Also, the code itself is not much use for replicating results without making the data available too. This can be more problematic politically and technically, but even here there are good precedents in the particle physics world.