![](https://secure.gravatar.com/avatar/69dcd277799bf98a2cd1c39dc48a82a3.jpg?s=120&d=mm&r=g)
While I don't contribute anything substantial to cctbx, a few months ago I ran cctbx tests after building Dials and wanted to make a couple of trivial fixes. I made a change and tried to commit it, but it didn't work - because I did checkout anonymously. So I spent a few minutes googling how to do authenticated commit from anonymously checked out sources, found nothing (I'm sure there is a way, but I didn't want to waste more time on it) and finally I copied the modified files to a laptop that I previously used to commit to cctbx and committed it from there. It's easier to contribute a change to a random project on Github. Not because the process in itself is simpler (fork, new branch, commit, push, pull request), but because it's the same process as for majority of other projects nowadays. Moving to Github lowers the entry barrier - for good and bad. Regarding boost, ccp4io_adaptbx, etc: I generally agree with Luc, but adding subtrees and submodules will make working with the repository more complex. No idea how it's mapped to svn for those who'd prefer to keep using svn. It'd be simpler to avoid subtrees and have ccp4io_adaptbx and other non-external modules in the main cctbx repo (it should be technically doable during the migration). Marcin