Even if they have the source, they may not have all the build tools anymore.

Or they have the build tools but the wizard that set up the build system back in the day no longer works there.

Or they have the build system archived and documented but it doesn’t run because some license expired, and the tool vender doesn’t sell that version anymore.

In the near future, there will be another possibility - SaaS cloud tools that are impossible to preserve so they are forever lost.

Wasn’t facebook also found to store images that were uploaded but not posted? This is just a resource leak . I can’t believe no one has mentioned this phrase yet. I’m more concerned about DoS attacks that fill up the instance’s storage with unused images. I think the issue of illegal content is being blown out of proportion. As long as it’s removed promptly (I believe the standard is 1 hour) when the mods/admins learn about it, there should be no liabilities. Otherwise every site that allows users to post media would be dead by now.

I’m fine with this. Instances shouldn’t proxy or cache images because it opens instance owners to a lot more liability than text. A client side setting to not load images in comments by default is better.

What you’re describing is no longer federation but full P2P. From a purely technical point of view, it may work, but the biggest problem will be abuse (spam, excessive resource use, illegal content). When a new instance shows up, how do you know if it’s a spammer or not? And if an instance is blocked by another instance, whose side should you be on?

This is true. If you run the reddit-grab project directly without using the warrior (sudo docker run -d --name reddit --label=com.centurylinklabs.watchtower.enable=true --restart=unless-stopped atdr.meo.ws/archiveteam/reddit-grab --concurrent 6 yourname), you can set up to --concurrent 20, and some projects do work well with higher concurrent, but not reddit. 6 is already pushing the limit.

I’m running reddit-grab on 25 VMs on azure (trying to burn my $200 free credit that expires in 10 days) and I can only run --concurrent 4 safely on most of them. The only VMs that can run --concurrent 6 are the ones in India, which seem to be soft-ratelimited by their higher latency anyway.

I think the biggest cost will be image/video storage. The text takes very little space in today’s standards. The good thing is that symmetric fibre internet connections are becoming more common so it may be possible for members of the instance to contribute unused disk space to help with its image/video storage. This plus limiting the image/video sizes (and maybe forbidding video uploads altogether) will allow the instances to scale with user count.