In the old days, software was shipped, then used, and eventually replaced or uninstalled. Academic publications are roughly the same: they get published, read and cited, and (for the most part) eventually replaced by newer results. However, while the way we create software has changed to where it now evolves, being updated regularly without uninstalling or fully replacing, publications have not. If you want to publish new results on a line of research you need to publish a new paper, even in cases where 80% of the theory or methodology behind it have stayed the same.
So I'm asking the question: does anyone know of a model where publications evolve over time, perhaps by being hosted and versioned on github or similar; where revisions can still get reviewed, but the authors can later change and extend their publication? Given the Internet, a model like that seems more appropriate to me. Publications could evolve over time from early position papers, to preliminary results papers, to conference style papers, to longer journal style reference papers, to books. Over the same time, the list of contributors could change and in particular, hopefully, grow.
Does anyone know of a publication model like that, either discussed in the library science literature or implemented in practice? I've looked at what some of the open journals do, but haven't found any model where publications can evolve.