[nodata]

Except if your files get corrupted, then you're screwed.

[r12e]

Of course, and the same is true of every backup solution. The way to mitigate that problem is with a decent backup rotation scheme:

https://en.wikipedia.org/wiki/Backup_rotation_scheme

[nodata]

> the same is true of every backup solution.

Rubbish.

A proper backup involves copies that cannot be modified after they are made (i.e. a snapshot after the backup has run if using disk, or for free on tape).

[r12e]

As it was ambiguous, I figured you must have been talking about the more likely scenario whereby files become corrupt on the live server, unknown to the either users or admins. Over time, these corrupt files then make their way through all generations of backup until no good copies remain.

There's nothing special about an rsync/ssh solution that precludes the backup server from creating a read-only copy of each and any backup.

[w1ntermute]

I solve this by supplementing the daily networked backups with weekly manual backups to an external hard drive that remains disconnected at all other times.