While theoretically true, the main bottleneck with Lemmy seems to be the database performance, so with both projects depending on PostgreSQL for that, I somewhat doubt that Piefed being written in Python will have much noticeable effect in reality.

The base consumer models from regular brands usually have no write cache and are thus cheaper, and obviously slower, but for data storage that doesn’t matter so much.

But you can also look into 2.5" HDDs if you are looking for power savings and noise reduction over regular HDDs.

The last one in the list.

The vote is missing one song.

Well, obviously if you host from your home ISP, people will be able to figure out your home’s approximate location via a reverse IP search.

But otherwise go for it. It’s not that hard to do and a nice learning experience.

This guy also being a perpetrator of bullying because he didn’t like moderation decisions makes this post a bit ironic though 🤷

This is odd because I know a few mainland Chinese people that use XMPP without problems (and afaik without a VPN).

Sounds like your server got blocked for another reason?

That’s quite competitive energy density wise to LiFePO4 batteries. I wonder what the prices are.

Well, instead of leaking metadata to Signal, AWS, Cloudflare, Google/Apple and your ISP, like Signal does, RCS only leaks it to your ISP /s

You can easily redirect xmpp to port 443 which is not blocked by most firewalls. If you have problems with firewalls or public wifis your xmpp server is misconfigured.

The actual military grade (xmpp based) messengers implement security lables, meaning messages are tagged with the required security clearance and if you invite random people to a chat they can’t see the messages.

Something like that could be also interesting for a Lemmy frontend to make it easier to share images on instances that have strict upload limits.

Hmm, maybe they are running some developer release? I don’t think this has made it into an official release yet, and if so that must have happend very recently (I tried selfhosting Pixelfed about 6 months ago and it wasn’t available yet).

This is a planned feature for Pixelfed that is perpetually finally almost finished since about two years now 😅

Well, like I wrote below, it is basically about scaling Lemmy across multiple servers, with all the complexity that entails. I am probably not the best person to given advise on that though.

If our smaller instance ever gets to a size that is not feasible to run on a single server anymore, we will likely close registrations.

I am not against showing cross-post links like Lemmy currently does. That helps finding communities that you have not subscribed to yet (but are already known to your instance). But automatically loading all the comments as well seems a bit too much to me.

We are talking about a feature that makes it easier to view comments, so general concerns about the feed concept aside (I find it a bad idea), it seems like forcing people to block communities so that their comments are not inserted into threads they are otherwise interested in seems counterproductive. It would be better to only show comments from communities that the user has actively subscribed to before.

Like for example I might be interested in the discussions happening on [email protected], but the comments in threads at [email protected] is something I would rather avoid. So this feature would force me to block [email protected] which is the opposite of making it easier to read the comments from communities I am actually interested in.

What happens if a thread is crossposted into another community that your instance knows about, but which you are not subscribed to?

I think it would be good to only display comments from communities you are already subscribed.

Lets agree to disagree on the “natural” centralisation aspect, which is IMHO nonsense. And very recently the US empire was beaten by some tribes in Afghanistan, so I think your argument needs some further thinking 😏

The reason it gets so much more expensive after a few thousand users is complexity. Up to that point a single server can be used and the necessary sysadmin skills are not very high. Basically anyone with a few weeks of training can rent a server and run such an instance.

After a few thousand users it gets steeply more complex, when you need to think about running a database cluster and load-balance the frontends etc. Not very many people have the necessary skillset for that, and even less are volunteering to do this. So you end up being forced to hire someone expensive with a high in demand skill. Basically your operation suddenly jumps from an easy to fund with donations volunteer effort, to a must commercialize or otherwise fund venture that is highly unsustainable in the short term.

As someone who runs a Lemmy server I can tell you that it isn’t as simple as that.

Yes, there is an initial benefit from having more users on an instance, but this initial scaling benefit isn’t linear. It rather abruptly stops at a few thousand users and after that it becomes much harder and more expensive to scale further. Only after going over that hump it might become cheaper again at the scale of hundred-thousand of users or so, but Lemmy the software is currently also unlikely to scale as a single instance to such numbers, so it isn’t just a system operator question.

So no, unless you want to fully commercialize the Fediverse and bring in external investors to fund the getting over that initial hump, semi-centralisation is not a feasible way forward. And what would even be the point of that? Reddit exists and is basically the same.

Luckily ActivityPub is designed to scale horizontially through lots of smaller (but not tiny) instances, so I think we can manage without the above.