Piero Bosio Social Web Site Personale

@b0rk generally yes but there are exceptions. Looking up what a specific argument does is fine in 100% of the cases, but rsync and sudoers are the two examples that come to mind when it comes to obnoxiously inscrutable structure, if you came to the man page to learn how to use the thing specific way.

@mos_8502 Essentially yes. Including also all the integration work, making sure everything links against the same shit well, handling slotted/collisions where necessary, having a package system, an install system, kernel configuration etc

For a very long while, Gentoo's internal "catalyst" tool was a widely used framework to make live images and custom appliance stuff in a way that didn't suck - I believe kaspersky used it for their livecd and appliances for instance

@flesh Sure! Decentralized implies some amount of peer to peer with no central authority at all, federation implies one or many central authorities _that I can have control over_, or self-host, where I have control over the data. Where it lives, what's done with it, how it is backed up etc.

It basically moves the "authoritative node" one level up in the hierarchy, instead of keeping it at the individual level.

Basically, I am ok with one or more central servers being part of the equation, as long as it possible for me to be in control of one, more or less.

@xgranade

@xgranade A great example of this is DeltaChat being essentially ideal at replacing the "group chat with friends" part of the equation, while being delightfully straightforward to self-host and demanding almost no resources.

It doesn't fit the "big public rooms" part just as neatly, like IRC ou Discord would, however, and I think that's ok.

I'm at the point in my life where I don't really need "decentralized" I just need federated with a hearty dose of data sovereignty. If that happens to come from a handful of good things, I'm ok with that.

@freya Sick! Thanks, that will give me a good excuse to boot these boys again <3

@freya That Could Run yes, that is running no, because it consumes the electricity of a small village lol -- but yes my Sun Blade 1000 was still in working order last I checked

I think my Sunfire v120 could also just be booted in situ, it is sitting on a shelf next to the rack in my lab

@freya I feel most of these, especially zones and ZFS.

I used to refer to Upstart as "Someone looked at Solaris SMF and thought 'you know what, I can fuck that up'", so maybe that is why I feel systemd is a step up.

Informal poll to see if there's some kind of correlation.

*If you have used Solaris SMF before*, how do you feel about systemd?

Went exploring old backups of old webroots and found this picture I took at work like, 20+ years ago

I thought it had been lost to time, but here it is, highlighting how JetDirect wasn't exactly a safe and secure protocol, and reminding me that the other admin got upset with me until I removed it

Fun times

@Viss I was under the impression the bridge was entirely third party?

What's the concern in this scenario, about them "just fucking changing their mind" (which they absolutely do) considering this?

EDIT: In case this comes across as the shitty "concern" kind of post, that's absolutely not my goal, I'm genuinely trying to assert the risk for my instance

@Viss @viq If your setup is very simple also, it is possible to skip vmagent entirely and have the victoriametrics server do the scraping itself, too.

@Viss @viq Yeah, or rather, you add the VictoriaMetrics server as a data source in grafana and it will read the timeseries from it. You can setup the data source __twice__ as prometheus AND victoriametrics (via the plug in) if you want, which lets you both do graphql and also just straight up use generic prometheus dashboards.

@viq @Viss You also straight up can set it up in Grafana as a prometheus source, it works

@Viss For reference also, it was all supremely trivial to configure. I put up a nuc-style PC with a N95 and 8gb of ram to hold the `graphs and shit` role in my homelab of like ~30 hosts, and it turns out that was WILDLY overspecced for the task. I don't know what kind of black magic VictoriaMetrics does in terms of storage and memory usage, but it's shockingly low.

@Viss I had similar concerns when I threw out netdata for Having Become Terrible(tm).

I didn't want to setup an elasticsearch cluster just to hold metrics.

It turns out grafana itself is fairly lightweight, and you can swap out prometheus server with something like VictoriaMetrics + vmagent, which lets you do push instead of pull depending on context, and also just normal endpoint scraping.

The datasource itself can either be transparently added as a prometheus one, or as a victoriametrics one (which nets you a few query language bonuses).

It's supremely low resource usage and does not consume much disk space either, to much of my surprise. None of it require a k8s cluster just to collect metrics either.

It's worth exploring, I think.