Though every time they release a new major version it breaks compatibility with the older major versions and this happens quite often. They need to stop doing that if they want mainstream adoption.
There seems to be one open source project that aims to implement Resilio Sync's model:
It's actually a true bonafide distributed systems problem, where the implementation and service matter just as much as the specification.
* When syncing a folder tree (uploading or downloading), it transfers files smallest to largest (globally across all files queuing to sync, regardless of folder). This simple algorithm is remarkably close to optimal behavior.
* When modifying a small file while larger ones are syncing, the small file immediately cuts to the front of the line (even sub-file). This avoids situations where a large folder uploading or downloading prevents you from getting the utility of the quick syncing of small work files.
* Files deleted from the server or another computer are briefly retained on the client in a special folder before being deleted from local storage so that their contents can be matched to other incoming files. In other words, even for very complex folder move and file rename operations, if the new files contain the same data as the old ones in whatever permutations, the client will sort it all out without re-transferring actual file data.
* There are no data loss holes. All operations are on a files-first basis (not folder-first), and Dropbox will never decide that a file was intended to be deleted unless the client app explicitly observes deletion. You don't need to worry that offline renames or moves, even multiple conflicting ones from different machines will result in deletions when the Dropbox clients go back online.
* "No data loss holes" also applies to conflicts of a single file. If a file is modified in two different ways by offline clients, one file or the other will be renamed to make sure that a modification isn't lost.
* The Dropbox client never gets stuck requiring the user to "resolve conflicts". It picks the sanest thing to do and does it (always observing the "never lose data" principle). Syncing is delightful and never fails, not a chore to be dreaded and carefully planned for with a bunch of "conflict" choices, like all systems prior to Dropbox.
There are plenty of open protocols for file sharing, like HTTP or SFTP. What unmet requirements do you envision?
Well, for starters, automatic synchronization in the background.
EDIT: Fixed typo