mastodon-archive is very useful, for Mastodon. We need the same for Lemmy.

Update: this announcement over the loss of endlesstalk.org reinforces the need for a backup tool. A good fedi server should not be fragile and dependent on the admin maintaining backups.

Update 2: This post gives some of the API calls that would be useful:

https://lemmy.ml/post/31128657/19022698

I'm probably not going to undertake this, but I'm considering it. Do you just mean download all your submitted posts, and attached images? If so, it seems very simple.

The mastodon-archive man page gives an idea of what it does:

Something that merely performs the backup function would be a great start, and the most useful function. But from there it could get quite complex with useful features. Mastodon-archive goes as far as to fetch content by others when it is a direct reply to something the user said. But I would favor going further and grab the whole thread of any thread the user spoke in.

Some non-essential but useful features in a Lemmy variant might be:

• Detect censorship on the user’s comments, which would entail a periodic comparison.
• Fetch new versions of edited posts - ideally use a simple version tracking tool.
• Find alternative links to the same post, thus tracking down cached versions on other hosts, and perhaps even detect host-specific censorship.
• A flexible restore mechanism, so a user could cross-post potentially in batches. So if my1sthost.net/c/foo vanishes, the posts could be reconstituted on my2ndhost.net/c/foo. (note that lack of data portability is attributed to someone’s decision to not use Lemmy).
• Shadowban detection: run as logged-in and also run cookieless, and flag differences.
• Non-Lemmy support (kbin,mbin,piefed)
• A MOVE option, which would make a local copy and then delete it from the server. This could get tricky because a user might want to only do this on msgs that have no replies.

In any case, a simple one-off backup would not need a huge effort. I suspect it would start with this API call: https://lemmy.readme.io/reference/get_user The natural next advancement would be to run the same job multiple times and expect it to not re-fetch already fetched content. For some reason that’s a little messy on Mastodon.. not sure about Lemmy. On mastdodon, Kensanada’s code decides that after seeing an already archived msg 5 times that it need not crawl deeper. Feels a bit ad hoc / non-deterministic.