The idea that research is "finished" and then published is flawed. This is particularly the case when we move from publishing results solely as narrative PDFs and instead publish results as compound, multi-stewarded and possibly reproducible Research Objects. Research Objects combine the discussion with data, methods, calculations, codes, scripts, workflows, platforms, links to third party services and so on. Methods evolve. Data changes. Metadata changes. Services get replaced. Platforms break. Stuff gets versioned. Things need repair. Lets face up to the fact that we release research rather than publish it. Lets start applying a 20th century software release paradigm instead of a 18th century print a book paradigm to scholarly communication.