Christopher Reay has demonstrated that wiki page json can be reasonably stored as files in a git repo without unreasonable growth as version compound.
Christopher has build a service that will watch registered site and increment backups as appropriate.
Andrew Shell has expressed interest in a similar approach as a follow on to his work on River. See Andrew's Backup Service Plan
Ward Cunningham was feeling guilty that we have lost sites larger number and more quickly than expected. Set out to duplicate Christopher's work. github
Are sites kept in their own repos or all stored as directories in one repo. Does one repo limit scaling?
Are repos checked out to local files to be updated or is one of several Github APIs used to update?
How does one recover an individual page, a complete site, or an entire farm?
What limits would Github or any other service place on transaction rates or total volume?
Would anyone expect assets to be backed up too? What to do with assets that are not found in Assets folders?