This question comes from not much experience with the bittorrent protocol, but what would it take to use something like this to build an open clone of btsync?

As pbnjay says, torrents alone can't be used because they target static content. I'm currently building an OSS alternative of btsync at https://github.com/rakoo/rakoshare.

Basically you can use a torrent to transfer content, but you need to transfer the new torrent itself some other way; I use bittorrent's extension capabilities to do that.

I also have very little low-level bittorrent experience, but the static nature of .torrent files would be tricky. You'd need some sort of snapshotting diffs to convert to .torrent equivalents at regular intervals, right?

Isn't syncthing https://github.com/syncthing/syncthing, supposed to be an open source alternative to btsync ?

It lacks some features (public shares for example) but works pretty well for personal syncing. It does not use torrents.

Sort of, Syncthing it's pretty nice.

I wonder if it would be easier building an open source version of btsync 1.4 as 2.0 seems to be a lot more complicated?

How mature is this compared to taipei torrent? https://github.com/jackpal/Taipei-Torrent

Shameless plug: I work on a BitTorrent tracker in Go https://github.com/chihaya/chihaya

What about for the server portion? I'd really like to distribute container images across a cluster of machines via torrent.

Have a look at https://github.com/chihaya/chihaya

TF run an article about it fairly recently. http://torrentfreak.com/private-anime-torrent-tracker-became...

I posted it here on HN before but it got no interest.

What server portion? The tracker?

Yes, I am thinking of an end-to-end library which you feed input files to and then it handles the distribution to all cluster clients.

BitTorrent is just the file transfer protocol, but you could very easily write such a library on top of it.