If you just want something simple that does the job, you can try a turnkey solution like YunoHost. There’s several other ones out there. Some with containers, some with more or less pre-packaged software… If you want to learn more during the process, maybe don’t and do it yourself because these things don’t teach you a lot. There’s some resources like the awesome-selfhosted list in the sidebar of this community. But I think for installing services you’d mainly look at the specific documentation of the specific service you’re just about to tackle. And maybe read up on Docker containers etc to judge whether you want to do it that way.

Haha. But seriously, what’s your problem? You don’t even use the software?

Of course. I’ve had a glimpse at the changelog every now and then and there’s a lot of substantial changes in there and multiple new features. I can understand how that makes it a bit more involved to get a release done.

Not sure if Lemmy admins are adventurous like that. From where I’m at, I can see exactly one instance running 1.0.0 code and that’s voyager.lemmy.ml

It’s difficult to discuss development in public places like here. There’s always a hundred things on the backlog. People want this, other people want that and someone needs exactly the opposite if all of it. There are a a bazilion ways users can annoy each other and all of it needs fixing. Then a project needs to be stable and reliable. It also needs new features. Performance needs to be right… It’s a proper nightmare job to balance all of it and maintain a mid-sized project. On top of it people will feel entitled, send in security vulnerabilities, complicated stuff that needs review and messes with things, other devs want something to be cleaned up, changed around, want someone to write more or less unit tests… and that also needs time for a plethora of good communication. And then there’s the actual architecture design and coding, which isn’t easy to begin with.

I didn’t study the code. But I’d bet the representation in the database stays the same, no matter which way it’s phrased on transport. It’s some sql relation between answer and post either way. A UI will also want to know how to style a comment at the point it processes that comment, so it makes sense to have it there. On the other hand it makes more sense for the semantics to have it attached to the post. Then there’s who can edit it. We need to trust incoming notes from third parties anyway. And maybe admins or mods can change it as well. They might be on arbitrary instances. So I’m not even sure if it changes anything with security.

And then there’s always many ways to skin a cat in software development. We can have long meetings to write specifications. We can choose to be a bit more explorative and figure things out along the way. We can even choose to make mistakes and fix them later. I think that’s a great thing with computer programming. Fixing mistakes is usually very cheap compared to for example a mechanical engineer who maybe likes to avoid wrecking a $1m piece of equipment. But that also means software developers have the opportunity to work a different way. And there’s a time for each of the methods. The trick is to apply the correct one at the correct time. I really can’t make a good statement here, I’d need to read the code and judge based on all the nuances I just mentioned. It’s regularly not as simple as something appears from the outside.

Ah thanks. I wrongfully assumed alpha meant alpha testing version. But seems there’s still a lot to do before that.

Good thing we have both development models and people can just pick what they like. I know which one I prefer. 😆 And seems Lemmy is approaching a release with their efforts of the last years as well, they’re at alpha.17 these days…

Yeah, I wish there was a good technical solution for this baked into the Fediverse. PieFed has some of that, you get some features for migrating communities there. But all of this is integral part of this place. We also have like 10 technology communities. It’s not obvious what to subscribe to. Some formed due to growth and changed dynamics. Some because someone was against AI and someone else pro AI, and they split off and made yet another community with the same name. None of that is intuitive to newbies. You can of course subscribe to all of them but then you’ll regularly get the same post 5 times in your timeline because it also leads to cross-posting and all kinds of things… This is by design, though. And it’s difficult to design online platforms to be easy to use, cater to all people, grant freedom to everyone… I think we still got some room for improvement 😉

Yes. Entirely different software. Different programming language and tech stack. Also different system requirements and feature set.

Not sure about the developer spirit. PieFed development has traditionally been moving crazy fast and it gets like several new features every month. I think that’s a matter of focus. It comes with consequences, though. But I think overall the project is doing a good job with trying to be compatible to other software. Prioritizing important stuff and doing the right thing. Sometimes some things get done, rather than be 100% perfect. But past experience tells me things often get fixed or changed around once necessary. Not sure if that’s a wise decision here. The JSON exchanged between the servers is probably extra work if changed around later.

I like getting updates and new features? My computer isn’t new by any means. But I tinker with stuff, sometimes bleeding edge technology. Other than that I don’t really care. Rolling release, Debian Stable… I’m fine as long as it does the job. And for half the stuff it doesn’t even matter. I can write a letter with a 5yo LibreOffice or answer mails with any version of the mail client. Just give me modern, up-to-date tools when developing software, and it doesn’t hurt if the slicer knows about my new 3d printer from this year.

Or maybe @[email protected] would like to do that for us?

Huh, Why is the 8B Ministral worse in the benchmarks than the 3B version? Nevermind, the bar chart is just bad.

This article is missing the point. Government should protect people’s rights. This isn’t a private matter. And companies are for-profit entities. They care for money. Less so for ethics and morality. And why is only sharing information with the government bad? Isn’t the real issue how they collect all the data about everyone in the first place? Is sharing that with privately owned entities okay? Like other people, your health insurance company, targeted advertising companies, hackers? I hope not.

Thanks. Yeah those would be great in an awesome-webhosting list. Or something concerned with household or businesses. But as far as I know you’re supposed to stick with a topic with those awesome lists and not make random lists of random projects… I’ve filed a bug report in the meantime: https://github.com/ccbikai/awesome-homelab/issues/24

Lol. Why isn’t Forgejo in Development but some predecessors are? And Gitea is listed twice. And why is a tower defense game listed under Automation? Also I think a few projects I use are missing. Why isn’t the most common content management system there? The second most common password manager? The reverse proxy everyone uses? And who on earth needs customer live chat and a lot of business-scale website analytics, webshop systems and CRM and ERP in their homelab?? I’m sorry but this looks like slop.

Did you try it? What kinds of results does it come up with? Usually if I tell AI to assume roles like those, it comes up with really over-the-top fiction stories. And what kind of framework and model do people use for pipelines like this one?

I’m not sure if this translates to the content creators. There’s many of them whom I really like to watch who do (or did) Youtube as a business model. Tom Scott being one example or Derek Muller (Veritasium). I’m subscribed to many more. Simplicissimus and their yet better second channel (in German). We wouldn’t have those without monetization. The platform of course went shit over time. Fortunately my Ad blocker still works and thanks to Sponsorblock my experience is fairly alright… But personally - I’m split on this question. We had quite the amount of entertainment before monetization but I think a large amount of quality content also arrived after that, and because of it. Those people would be working some office job today if it wasn’t to Youtube. And I (and the world) would miss out… On the other hand we got MrBeast, a lot of fake cooking videos…

One thing I did is connect to the smart home (Home Assistant) and the NAS running at home. Some internet service providers don’t provide proper IPv4 addresses any more so IPv6 is the most convenient way to connect. This doesn’t require a VPN provider, though.

As far as I know it uses the B.A.T.M.A.N. mesh protocol. On a channel within the regular 2.4GHz wifi spectrum. So no license needed unless it collides with laws for point-to-point beams. All people communicating to each other obviously need to agree on a channel. It comes with some hierarchy where I’m at. There are local chapters who make up some config and who also operate nodes and exit nodes into the internet. These are necessary because Germany has stupid laws.

There’s Freifunk as well!

Don’t MoE models just load into memory as every other model and it’s just that they pick a subset of numbers to multiply by each step so they’re faster? That’d make me think it’d need somewhere around 80GB of memory at 8bit or 160GB at full precision, or something like 50GB at the average llama.cpp Q4_K_M…