Thanks, the one I’m currently using is not updated to 0.18.3 yet

If my fingers prune I’m going to die or something

It seems to work, but it keeps throwing pictrs related errors, so it’s not really built for it

It barely works on my old version of lemmy, probably fixed now then. It would be nice if there was a was to turn that off and only use pictrs only for locally uploaded images. Since I’m the only person here caching isn’t too important.
I wonder if I could shut pictrs down and only use an external image hosting for images?

Lemmy sometimes caches remote content in pict-rs. It’s a bit broken so you usually don’t see it, but it does do it occasional

Related, What about a personal instance only I use? I can choose what communities I want but I can’t control what is posted on those communities. Someone could post something illegal to a beehaw community (and have) and the mods remove it, but does the deletion of images and posts federate? In know matrix keeps copies of every deleted file in a room on all homeservers, what about lemmy?

I have done some testing and I found a few reasons I’m having issues with webtorrent:

• The only reason they where working at all is because they were downloading from the HTTP URL in the torrent file, P2P was not working at all.

• To download the webtorrent from the blender instance I need to have the video watched in my browser to peer with the webtorrent client, the instance peers don’t work on non-peertube webtorrent clients.

• The reason instant.io was broken is my adblocker was blocking the tracker.

• The tracker in my peertube instance is broken.

EDIT:
I was a bit wrong here, there are two different formats in peertube: webtorrent and HLS. I was getting confused why the video on my instance (HLS) and the one on the blender one (webtorrent) was behaving differently with webtorrent clients. They are completely different formats so that makes sense now.

Webtorrent seems to have some issues with peer discovery. I’ve tried the instant.io site they have linked on webtorrent.io and I can’t get it to download or share anything, the desktop client managed to download a torrent from my peertube instance over normal BitTorrent but I can’t share it over webtorrent. I downloaded a video from my peertube instance using btorrent.xyz over webtorrent but I can’t seed new files because the peers don’t find each other. when I use a webtorrent with a tracker (like peertube) it works fine but how were sites like instant.io supposed to discover peers without trackers? I don’t think DHT exists for webtorrent yet.

You can manually seed videos on instances using redundancy but I was thinking automatic redundancy for watched videos might be a good idea, I guess you can do automatic redundancy for entire instances but that would take up a lot of storage space.

One of the nice thing with BitTorrent is the high reliability so I assumed that was what peertube was trying to do, I guess the idea is not to provide data redundancy but to split load instead?

why? if 5 instances are seeding the video, clients should be able to download from all 5 instances and spread the bandwidth usage right?

Why not also use the instance to re-seed? it could keep seeding after the visitor closed the video.

Would it not make more sense if your instance downloaded and redistributed the torrent? then you could keep seeding after the tab closed. it also wouldn’t leak your IP then.
What about peer discovery? I opened that webtorrent website in two browsers and they didn’t peer, is that demo real?

This is wrong, I use IPTables but the device is absolutely not dedicated lol.

Download ML thing.
make new venv.
pip install -r requirements.txt.
pip can’t find the right versions.
pip install --update pip.
pip still can’t find the right versions.
install conda.
conda breaks for some reason.
fix conda.
install with conda.
pytorch won’t compile with CUDA support.
install 2,000,000GB of nvidia crap from conda.
pytorch still won’t compile.
install older version of gcc with conda.
pytorch still won’t compile.
reinstall the entire operating system with debian 11.
apt can’t find shitlib-1.
install shitlib-2.
it’s not compatible with shitlib-1.
compile it from source.
automake breaks.
install debian 10.
It actually works.
“Join our discord to get the model”.
give up.

Why does The Register get paid? /s

I have a stupid overcomplicated networking script that never works. So every time i set up a new server I need to fix a myriad of weird issues I’ve never seen before. Usually I setup a server with a keyboard and mouse because SSH needs networking, if it’s a cloud machine its the QEMU console or hundreds of reboots.

The file you downloaded is a compressed JSON file, it’s not something you can really just look at. But it contains all the data needed to build a nice UI around.
I don’t know what OS you are on but on linux you can run zstd -d -c file.zst | jq . and it will print everything in the file. It’s not really readable though. Also it doesn’t have any of the media content, only the text

I hate reddit. But it feels like the library of Alexandria burning down (yea I know). All those google search results and educational subreddits that are shutting down forever, and because they are too small reddit won’t force open them again.
A lot are in the pushshift archive, but that cuts of at 2022. Also, it doesn’t include a lot of the smaller subreddits.
I have had my PC running 24/7 with multiple VPNs to avoid rate limits downloading as much as I can before the API dies, but with some blackouts moving forward a day I have already missed a few.
Like many others, I would often add “reddit” to the end of my searches to get better results, half the websites on web searches now are either AI generated, copies or are completely AD ridden websites that ask you to turn off your AD blocker.