Very smart to pause and weigh things. There was a lot more to it than I thought when I started mine.
- Renting a server is not cheap, but it’s not expensive either. Most VPS’ will be able to host it fine. You will need to invest effort into swapping data over to S3, volume storage is what really costs money if you use it incorrectly.
- I’ll say do you want to have an individual/known user instance (like family and friends you deeply trust), or do you want to allow randos? The thing you don’t mention at all is liability. I don’t know where you’re hosted but in most countries you are responsible for reporting material on your server. CSAM? Guess what, you are responsible. You can look at my server, it’s based out of the USA and any server is automatically a “mandatory reporter”. I am legally obligated to report CSAM that makes it to my instance, if I don’t I am legally complicit.
- Note that this has been drastically reduced with the image proxying, where if someone on say, .world posts CSAM it’s proxied through my server but not hosted by my server. So, liability is still a thing, but as long as the admins of .world take action then i’m protected with them. If proxying is disabled then the CSAM would live on my server too - and that means I’m legally required to report it.
- NEVER allow open signups. Spam is real here on the fediverse, and bots are actively trying to sign up for accounts. If you allow signups you need to require a captcha, and I recommend either email verification or asking for a signup message “Why do you want to sign up?”. This gets rid of 99.9% of spammers.
My suggestion is to start a personal instance first. Get the feel for it, see how you like it. Maybe create one community on there that you’re passionate about and advertise that it’s there to the fediverse (since they won’t know about it until you tell them about it). Then judge your risk level and see how much you’re willing to do. For me, I host a bunch of swifties, it’s well within my risk tolerance. I approve everyone that comes in, and most have to ask to join.
None of this is meant to scare you off, obviously I still host and I’m glad for it. My Swiftie Community has over 1,000 subscribers now! I’m very happy to host our little niche community, but I also have learned a lot on the way.
poVoq@slrpnk.net 4 days ago
I am not sure OP is asking about hosting a Lemmy instance though. They mention non-Lemmy fediverse software.
However I think you misunderstood how proxying works:
It is actually pretty much the oposite of what you describe. The image proxy in Lemmy is a user privacy feature, but it comes with the downside that the server does indeed download and temporarily stores all media that are requested through it.
scrubbles@poptalk.scrubbles.tech 4 days ago
It’s halfway between us. Without proxying images are pushed into my server and I end up hosting them indefinitely, requiring me to manually review and remove something if it’s removed on another server. (Moderation actions like that from what I understand are still not federated, although maybe that has changed in one of the last updates).
The proxy feature is a privacy feature, but for us admins it also works from a liability standpoint. If proxying is set to
ProxyAllImages, it will send the image URL down to pict-rs. From there, pictr’s will cache the image based on time that you set. So yes, it’s stored temporarily in Pict-rs for quick retrieval, but then I time out after a day also, so if something was banned I hosted it for max a day before it was purged. It removes me manually needing to manually trace back to events that happened a month ago and wiping it from my S3. If my S3 was searched you’d find images from today, and a bunch of images of Taylor Swift.Shadow@lemmy.ca 4 days ago
We dont proxy on lemmy.ca yet, but I assumed thumbnails would still be stored in pictrs like usual? I thought it was just the actual image links that got proxied, and thumbnails were still dumped in like usual.
We turned on cloudflare’s CSAM scanner and remove anything it flags for us.
scrubbles@poptalk.scrubbles.tech 4 days ago
They are but there’s an environment vatiable to set how long it stays in cache. Cloudflare actually just ended their csam auto submission, at least the auto reporting. Does it still at least flag it?
poVoq@slrpnk.net 3 days ago
Hmm, afaik other than some generated small thumbnails no remotely sourced images are stored on your server when you turn off the proxy. At least in theory, but the entire Pictrs integration in Lemmy is such a mess with random unexpected behavior that at this point I am hesitant to claim that no remote images ever get stored (there seem to be alternative code paths for specific image hosts like Imgur and crap like that).
scrubbles@poptalk.scrubbles.tech 3 days ago
I read the source code, the proxy essentially just hands all the responsibility down to pict-rs. Then pictrs has that environment variable where you can set how long-loved you want the thumbnail to be.