Not disagreeing with you, but here are some ideas on how github could take things to the next level. 0. Get Junio Hamano from Google and several of the top git

Not disagreeing with you, but here are some ideas on how github could take things to the next level. 0. Get Junio Hamano from Google and several of the top git-core contributors, so they can start releasing open source patches to git that support github's long term roadmap. 1. Come up with efficient custom diff functions for a variety of common file types, not just image diffs but also dbdiffs, mp4diffs, and the like. Github has already done some work in this area, and a little known fact is that git already has some support for custom diffs in .gitconfig: https://github.com/blog/817-behold-image-view-modes http://man.github.com/git/git-difftool.html http://stackoverflow.com/questions/9154110/data-versioning-with-git-custom-diff-for-specific-files 2. Next, integrate git-annex for storing huge files in git into git core, maybe with a configurable option for where to place the files (perhaps on a github-customized API to S3). http://git-annex.branchable.com 3. Now, with some thought and work, you could start supporting diff-based workflows in a wide variety of disciplines, including movie, image, and document editing. 4. You might also start expanding the range of github to start including db versioning capabilities. I am convinced that the successor to Postgres will include schema versioning as a first class capability. One way to do this is to allow every table to have a list of historical schemas s_i each associated with the instances v_j that were populated when that schema was in effect. Then all you need are functions that map from s_i to s_{i+1} to migrate historical data forward in time.