Prevent or warn when changing the synchronization method or server

[locnide]

Currently, when you change the sync configuration (for instance switching from local to server), there is no warning, and nothing prevents you from syncing.

But it seem that it will only sync the operations not marked as synced, thus making the data in the new sync backend incomplete. A newly synced device using the new backend would miss operations, wouldn’t it.

---

I have four proposals (I didn’t look into the taskwarrior internals, so far)

1. Forbid (or warn) sync when the method or config has changed.
   The implementation could add a sync_config key that contains a footprint of the sync method (server+URL, local+DIR, …), and refuse to sync without force if it has changed.

2. Add an official (subcommand), or unofficial (script/documentation) method to "unsync" your local replica : it would set synced to 0 for all operations and remove information in sync_meta . It might be a confusing feature, though.

3. docs Remove Saves disk space. from the documentation. It seem that all operations are stored in the taskchampion db. Are some operations removed from the local db when syncing ? edit still relevant, see comment below

4. docs Add a warning in documentation to explain that changing synchronization method or configuration is not supported. It could be an easy step before 1. is implemented.

[djmitche]

Yes, for a while operations were deleted immediately after they were sync'd, but in #3671 we started keeping them around as long as the task exists, as a kind of "history" of the task. However, once a task is purged from the task db, its operations are also deleted. So it saves less disk space than it used to, but that sentence is still accurate.

It's "supposed to" work to move from one sync method to another. In fat this is the same issue as #4123, which has an associated Taskchampion bug. The idea is that the sync data (in the form of a sequence of "Versions")) remains the same, but some of it is missing on the new sync server. Snapshots are what make that work: when an empty replica tries to sync against the new server, it downloads the latest snapshot and then all versions since that snapshot. However, as in #4123, if there's no snapshot then this doesn't work. So, we need to have a way to signal the urgent need for a snapshot in this situation.

As I mentioned in Discord, it might be that this is just so difficult to get right without understanding the protocol, that we should mark it as unsupported, and suggest an export and re-import as a workaround.