Geoffrey Goodhill questions some of the practicalities of open data-sharing policies (Nature 509, 33; 2014), but I believe that his concerns are largely unfounded.

Storing large volumes of raw data is costly, but many items destined for sharing are highly processed and relatively small. The mouse-brain connectome, for example, is available as a 3-megabyte file derived from many gigabytes of raw data (S. W. Oh et al. Nature 508, 207–214; 2014). Neither is there a shortage of repositories: many institutional databases are freely available and well supported (such as zenodo.org, maintained by CERN, Europe's particle-physics lab in Geneva, Switzerland). More repositories will come online as researchers learn how to share data more effectively.

Contrary to Goodhill's suggestion, sharing computer code does not necessarily demand much time investment (see, for example, D. C. Ince et al. Nature 482, 485–488; 2012). Code is a valuable part of a paper, so everyone benefits if its authors assume from the start that it will be shared or reused. Also, people releasing code are under no obligation to maintain it.