I have been on and off Borg for years. Mostly have used the vorta GUI which works OK except when there is a python problem. Which you may or may not notice when it happens.

Pro:

- the deduplication is superb. First time I ran it I thought I had misconfigured badly because the final backup was about 1% original size. But this is what happens of you run it on a bunch of uncompressed files like text and other nerd shit. And it hardly increases over time doing the incremental updates. The 10gb free from borgbase.com will backup all essential configs for the rest of time for all devices.

- a bit more complex and need to do it via CLI but you can even dedupe across machines to be very efficient

- even tho the price per GB is more at Borg base compared to other off site storage, IMHO this is offset by the efficiency

- using a specialized server let's you offload some of the work of making the backups from your local machine so running it all the time is less demanding of resources

- lots of nerds use it and that gives me confidence.

- you can use something like rsync or whatever (even syncthing) you like to make a local backup 1 version, then use Borg to send it off site (or to your own server whatevs). All histories will be deduped, pruned etc so very small footprint.

- Linux, floss etc. No electron or other annoying BS.

Cons:

- the actual CLI application is very complex. After trying what is available I think emborg is the best front end for me. You still need to write a bunch of scripts and otherwise fck around. Anybody who describes Borg as easy, foolproof etc is a damned liar. It is a powerful tool that requires you invest in it.

- the documentation is very long. The authors confuse length with thoroughness. Explanations go on for pages, but then give only small simple examples for the most rudimentary use case. So it is hard to grasp the concepts. It is so long and sprawling you get lost in it. Especially when you add the secondary docs of front ends, wrappers etc onto it. You remember reading something... Somewhere.... but then can't find it again.

- case and point of the above: inclusions and exclusions. Key component of backup i think. There are multiple systems going on at the same time. The systems are similar and different in ways you are meant to infer from ??? (your computer science degree??). After hours and days trying to work out how to make a pretty normal system backup, I started to think it might just be a weakness of the application rather than purely user error. I found many threads on reddit, forums, github etc of people with same basic questions never getting a good answer. There also isn't a good way to test inclusions and exclusions.

- putting data in is one thing (easy) getting it out is another. What you need to accomplish this is VERY unclear especially if using the GUI but also in the CLI. In order to test retrieval you must do it on a totally separate machine or you may be falsely reassured. I have lost data due to this, only an inconvenience as it happened but I never feel totally safe with only Borg to rely on. Again I don't feel it is me being stupid because the internet is full of long threads of people trying to wrap their minds around how this is meant to work. I honestly still do not understand how passwords work, and I'm afraid if I change any password it will break everything. Which isn't great.

- from how I have understood it, its strongest on retrieving for example 1 document or 1 directory you accidentally deleted. I'm sure its possible to restore an entire system but I wouldn't bet too much on it being straightforward. Would probably be better to use an intermediary system backup tool rather than Borg directly.

In conclusion, Borg is as it appears: a powerful tool built by nerds for nerds.