Protocol - Retransfer of the index - based on actual change

Development
1

Hi!

When Syncthing is restarted for config changes or some other reason, it is always retransferring the index to all other node.

All nodes know:

  • When they last saw the all configured nodes.
  • When was the last file change

If there would be a short handshake:

  • Did you change any file since you last saw me? Answer yyyy
  • I have xxxx changes since I last saw you. If the answer to xxxx and yyyy is “0” then the index should not be retransmitted.

That would speedup the time to sync a lot. Especially on nodes with large repos and many nodes. And also makes the frequent restarts less painful.

2

Yeah, it’s just not implemented yet. (See the small section about “Max Local Version” here.)

3

Thank you for your answer, this was excatly what i was looking for.

Should i make an issue for that on GitHub?

In the “Fields” section there is a line: Exactly one of the T, R or S bits MUST be set. I didnt see the use or the description of the “S” bit. Is this bit obsolete?

4

There is now a GitHub issue to this: