Well I'd say that a lot of people will agree on the principles of reproducible research and will clearly see the benefits, but I am surprised no one has mentioned yet how challenging this is in practice for neuroimaging. First there are huge issues with anonymization, especially for clinical data. Processing a dataset is one thing, releasing it publicly is another (faces in T1 scans have to be blurred for example). Then you need to host securely tons of datasets online. In the same vein, coding an in-house algorithm is one thing, releasing it publicly is again completely different (you need to document !). For new algorithms, the production environment can be very hard to reproduce. Moreover, the analysis itself can be computationally challenging (I use supercomputers all the time, this is not used by the vast majority of the neuroimaging community). Not to mention that a lot of research group do not fully automatize their data processing flow. My point is that it is not enough to say "let's go reproducible/public/open source/....". We also need an infrastructure to do so. I bet that in the next couple of years we will have websites that allow to share datasets publicly, and request with only a couple clicks for processing in supercomputers with well tested and maintained analysis pipeline. There are many current efforts in that direction (e.g. http://www.cbrain.mcgill.ca/). But at this stage this is science fiction as far as I know.