nice to know! Please, continue posting new things in modern c++ )

It can be independent only if you “own” hardware/software, i.e. you can manufacture, change etc. If there is no hardware independency – you loose with software, then.

What is the need of that if there is a std::pmr::vector and similar?

Ha-ha! Most of people thins that syncspirit UI is terrible :)

Well, yep. For now the positive differences are

• better speed transfers (up to x7 speed up on my setup).
• different UI (including unique features like see in realtime tree-like structure of remote/local folders).
• Windows XP support
• it targeted for long-awaited “selective sync” mega-feature (asked for many years by syncthing users; but it is out of scope of syncthing business, so, it will unlikely be developed). I plan to do it after 1-2 releases (after implementing “ignoring local files” feature).

Did somebody actually tried it? On real hardware? How it is?

Dinit for desktop, s6 for server.

What is wrong with s6 for desktop?

Not yet (I’m a solo dev so it is quite hard to do multiple builds). Patches welcome! :)

Somebody might find that behavior a bit making nervous if you don’t see progress, that it is running or that it was synced a few mins ago…

That’s why for that “daemonic” behavior programs hide themself in system tray :)

I think it will always stay compatible on protocol level.

That tools does what I always think was possible to do, or am I missing something?

I think the problem is that they do not share the whole communication protocol/model with GUI/client and it has only feature to block something aposteriory masks, while the whole folder with all files is already shared with client’s device. The original syncthing database scheme also seems does not supporting this.

Technically, from a protocol view level, there is no problem just to ask a single file to download and share only it.

I think because of that they do not implement “selective sync” feature, i.e. when you need to sync only a slice of remote directory.