3 ms·

it's an interesting project! you may want to take a look at git-annex as well, which also uses git and ssh (and rsync) to synchronize things but also supports l

• dantillberg 11y ago
  Thanks! The switch from Node to Python was mostly incidental -- I had written a peer-to-peer style program in Node, and I wanted to simplify the logic by rewriting it with a primary/secondaries model. The switch from Python to Go was more intentional. While running gut-sync, there are times that sub-commands that can take a long time to complete, e.g. git-fetch or implicit calls to git-gc, and git provides feedback in the form of progress meters. I wanted to be able to have gut-sync do simultaneous fetches to N servers, and to have statuses interleaved in the console in real-time (I'm doing a bit of TTY hacking here). I tried rewriting logic first using Python 3.4's asyncio, but I found it too complicated and too slow to do non-blocking I/O in Python. There are readline() methods and read(N) methods, but I had a ton of trouble getting them to work in non-blocking mode. Basically, Python's subprocess stuff works great if you're fine using proc.communicate() or proc.stdout.readline(), but if you want to do something real-time, the lower-level stuff gets very complicated and slow. asyncio is a step in the right direction, but it was still pretty awkward to have to specify `@asyncio.coroutine` and `yield from` everywhere, and it could be fairly tricky to track down whenever you screw one of them up. Over in Go-land, though, you can just read streams one byte at a time if you want, and do whatever sort of fancy processing you need to, and it's all blazingly fast by comparison. And go-routines are very simple and flow naturally with the language (compared to feeling bolted on like an after-market mod in Python). Edit: PS I still love Python. It's wonderful for many programming tasks. It's just not as great for low-level real-time systems programming as Go.