

A fair amount of drama is exactly their fault. Mozilla chose to increase management pay and fire people, Mozilla chose to flirt with ai, Mozilla bought an ad firm, and so on. It’s not like someone was holding a knife to their throat.
Blog op doenietzomoeilijk.nl.
A fair amount of drama is exactly their fault. Mozilla chose to increase management pay and fire people, Mozilla chose to flirt with ai, Mozilla bought an ad firm, and so on. It’s not like someone was holding a knife to their throat.
Yep, the M is for mb, that’s for RAM in this case.
As for levels… That’s not really that black and white, IMO, there’s no “best” platform, it’s always “it depends”. I think you’re fine with a pi, certainly for a while, and especially at the price point. Plenty of folks running fedi instances and Matrix off of 'em.
The only thing that comes close and has good software support would be a second hand small form factor office PCs, like HP mini PCs or the Lenovo ThinkCentre. Might buy you some expandability down the road, but it’s slightly bigger, uses a bit more energy, choices. It depends.
To start with the last question: yes, you can absolutely host more than one service on a single machine, resources permitting. The different services will each listen on a different (TCP) port, and you can front it all with a proxy which works a bit like a front desk, directing the incoming requests to the proper port, so foo.example.com gets directed to service A and bar.example.com gets directed to service B, and so on.
The key part is “resources permitting”, because all those services need CPU cycles to run and memory to run in (not to mention storage). Especially RAM is critical; have too much running for the amount available, and your server has to “swap”, parking bits of ram to disk, use it for whatever has to run at that moment, and swap bits back. Storage is always vastly slower than memory, so this slows things down tremendously, to the point of the server feeling sluggish or frozen. If you run on a Pi that runs off of a microSD card, not only is your storage really, really slow, you’ll also severely limit its lifespan with swapping. So do invest in better storage, like a USB NVMe drive (not a regular USB thumb drive, as those are typically the same flash storage as sd cards). And see if you can get a pi with more RAM. There’s no such thing as having “too much RAM”.
So, what to run? I don’t know about Hubzilla specifically, but their FAQ (under the “average hosting cost” header) says you should be fine — it’s just a PHP + database app. But with apps like these, it also depends on the actual use: if your family and friends start following a million people, that’s going to increase resource use. Keep in mind that over time, you’ll see storage increase slowly but surely, anyway, I’m running a single user GoToSocial instance for myself, and the database and cached images and whatnot amount to some 12GB of storage. I did mention getting extra and faster storage, right? ;) I know there’s folks hosting GoToSocial and snac on severely constrained hardware, like raspi zero (so far less powerful than what you have in mind), old routers and even their car radio…
WordPress is just another PHP + database app, although it tends to scale somewhat shitty; if you’re not entirely tied to WordPress, you might look into different systems, maybe a static site generator that turns your pages into, well, static HTML files, which take next to no resources (CPU/RAM) to host.
Synapse is a bit of a heavy thing (although it has gotten vastly better, the last couple of years), but it too is quite disk-heavy, so really don’t run this on SD cards.
Point is: yes, you can absolutely start with a Pi. I’d try and get one with as much RAM as you can / are willing to spend, as you can’t upgrade it, and get some storage that’s faster and less prone to failing. But even 2GB will get you some way and you’ll learn a ton (aka “break stuff”) in the process!
In Dutch, we have the similar “zoals de award is, vertrouwt hij zijn gasten” (roughly “the way the innkeeper is, is how he trusts his guests (to be)”).
I mean, yeah, this is pretty easy to toss into my backpack.
I have a slightly bigger board (a Lily58 that I built earlier) that lives permanently at the office. I occasionally use the regular laptop keyboard, just to keep that bit of muscle memory, and switching is usually pretty easy.
Full size boards look weirdly big, though. 😂
Wow, some people will just not hear about living without their ISO enter, huh? 😉
I believe they’re absolutely not street legal in the UK, nor in the EU. Those were never “ridiculous sized trucks” Walhalla to begin with (although I see more Rams than I care to, these days), so there’s roughly zero chance those things will become mainstream here.
Heck, we have rain here, that’s enough of a wankpanzer repellant.
This is a Sweep with choc switches and Nice!Nano knock-offs for controllers.
Actually, it’s not that expensive in the grand scheme of things, I’d say about €65-ish. That’s the PCBs (the electronics prints that you solder the rest onto), controllers, switches, keycaps (both relatively expensive because they’re low-profile) and batteries. The schematics are open source. If you want to start cheaper, build something with MX type switches, rather than Choc switches, you can find both switches and caps quite cheap. Or, if you don’t want to play “hunt the part on Ali express”, there’s companies that sell pre-collected (and sometimes even pre-built) kits.
It takes getting used to, of course, but at least for me, it quickly became second nature. So no, I don’t miss having more keys, in fact, having a num pad right under my right hand (rather than having to move my hand and arm to the right) is quicker as I don’t have to find the right spot twice.
As for quicker… I type about as fast on this as I used to on a regular board, but this is more about ergonomics and comfort than about raw speed.
For what it’s worth: I’m a developer-gone-sysadmin, so I spend a decent amount thinking and/or cursing computers, typing is only part of the job. Plenty of IP addresses, though, so I get my numbers in. There’s some documentation and blogging as well, so long form text.
It is indeed!
The tech might be the same, but the models can certainly differ, and something that is trained on US-centric data gives US-centric results, which may not always be desirable.
Yeah, no, not anybody can host a server. Sure, you can host a PDS, but the AppView still wasn’t open source last time I looked, and hosting a relay requires tens of terabytes of storage, not to mention the bandwidth to keep up.
Meanwhile, people host actual activitypub instances on repurposed routers and their car entertainment system…
…I have no words.
They can’t even do a coup without fucking half of it up.
From boiling water into water that happens to be in a switched-on kettle. Huge improvement.
When the enshittification comes (when, not if, they’ll have to drag their feet to move somewhere else again. All their followers will have to follow them again. Had they moved to a proper open solution, they could’ve stayed there indefinitely.
It’s not just about Bluesky" not being the proper and pure solution", it’s about this being a temporary measure at best, and people don’t seem to realize that.
If I understand this document correctly, it would mean that the entire connection somehow gets routed through Meta’s servers. I can fully understand the reluctance of other parties, including Signal, to do that, and I wonder how this is actually compliant with the DMA.
Neil doesn’t need a chatbot with sparkles for that, he’s plenty capable to take absolute piss himself. 😁