What if copyleft was applied to scientific publishing?

Copyleft is a big success because reciprocity creates a snowballing effect. For example, if you take Linux, the entire source code of which is freely available under GNU GPL, make some modifications and distribute the derivative work, this derivative work itself must be licensed under GNU GPL. Therefore the source will remain open, and improvements effectively become available to public at large. Very soon, it becomes too costly to maintain a private closed-source fork of a copyleft software.

Scientific publishing is going through a similar transformation. There’s a movement to move to Open Access for publishing research papers so that works are available to anyone without requiring any subscription fee. However, things aren’t progressing fast enough:

A history of publication in prestigious journals is a prerequisite to every step on the career ladder of a scientist. Every paper submitted to a new, unproven OA journal is one that could have been published in heavyweights like Science or Nature.

Governments mandating that publicly financed research be made publicly available is one way to get out of this catch-22. I believe a second way might be to incorporate reciprocity in the OA license itself.

What if Open Access came with the requirement that all derivative works — research papers that cite or build upon the previous one — are themselves required to be under the same license? Thus, OA works could cite and build upon all kind of works,but closed access papers could not take advantage of OA works.

One immediate wrinkle in this scheme is an actual legal implementation of it. Copyleft uses the copyright law for enforcement. It’s the act of copying that is prevented by law; copyleft allows copying, but with a restriction. Citation, on the other hand, is not restricted under current laws at all.

But let’s assume that there’s a legal breakthrough equivalent to what Richard Stallman achieved with GPL. Could it work? Could this do for scientific publishing what copyleft did for software? Or instead, would we be harming ourselves by disallowing closed-access derivative works?