Things that frustrate me immensely: There is no hard requirement for equipment like heatpumps and EV wallboxes to be smart in any way. No law that sold vehicles must be able to feed energy back to the grid. We are effectively building out infrastructure we know to be "obsolote" like this. The future will be EVs trading power on the grid, charging when it's plentiful and supporting the grid when it's harder to generate. Heatpumps need to be able to dynamicallly reduce their load on the grid in case of a need of power - and people need to be rewarded for being flexible like this.

Frankly the current state of this technology is a joke. Wallboxes are as smart as a bad calculator (at least *some* have the support for extending them via RS485 or similar, while basic models are basically glorified power outlets), I have seen a new heatpump without any smart external control. It needed a 400€ module to even speak RS485, it couldn't take into account weather forecasts or anything like that. And it's not a cheap system. German name brand, as expensive as a small-ish car.

I am absolutely astonished how badly this is regulated. It's as easy as making it a law - you want to sell a wallbox? It has to have the ability to receive updates and to take control via LAN (fuck this cloud IoT rubbish, it needs to work locally with a local energy management controller). You want to sell a vehicle? It has to be vehicle-to-grid ready according to whatever standard we enforce. You sell a heatpump? Same as the wallbox - it needs to have LAN connectivity for local energy management. Also IMHO this must include user-driven updates, without a manufacturers cloud and some form of escrow - if the company making those products goes bankrupt, the documentation of how those products work internally must be publicized so they can be maintained and updated. We also need well-enforced standards for integration to local energy management hubs.

All the glory of my magic headscale setup. The reverse-proxy serving http://statuspage is running on my central control server, the statuspage itself is hosted on my home assistant raspi at home.

The next step could be adding automatic LetsEncrypt support via a DNS-01 challenge, as my tailnet uses a completely valid domain that I could get validated certificates for - but that doesn't really matter much to me, given I'd rather use the short hostnames made possible by a correctly set search-domain and it's all encrypted with wireguard anyway.

While I was pretty happy with my headscale/tailscale setup, I missed one part: Vanity hostnames for services. In my tailnet, only machines get nice hostnames, but for example I want access to a host like http://grafana/ for my central Grafana (a service on a random port on one box). It took a total of 49 lines of go code to build a reverse-proxy that does this. This proxy is it's own node on the tailnet, therefore it gets a hostname and takes part in the usual routing. It then uses the reverse proxy lib built into golang to proxy to the real host.

The algorithm is proud, I guess.

He talked about electric cars. I don't know anything about cars, so when people said he was a genius I figured he must be a genius.

Then he talked about rockets. I don't know anything about rockets, so when people said he was a genius I figured he must be a genius.

Now he talks about software. I happen to know a lot about software and Elon Musk is saying the stupidest shit anyone's ever said, so when people say he's a genius I figure I should stay the hell away from his cars and rockets.

==> Escalating privileges using sudo

==> finished: 145 packages removed (disk space saved: 853 MiB)


Arch Linux update rates are a bit staggering, I updated this computer like... last week.

@kleinsophie That looks like you are missing the WindshieldController configuration, no control-loop to reconcile the defective windshield seems to be active!

@roland Aaaaaaahhhhhh :D

What I'm lacking: iOS support for Headscale so I can take my Home Assistant instances off the public internet, that would remove the need for CloudFlare tunnels to my parents (fucking carrier grade NAT prevents me from running this in another way). Technically I have enough bandwidth and compute at home to get rid of the VM, too, but it's just too damn convenient to have that on a static IP on the public internet, compared to my changing home IP.

@roland True, but I'm running Akkoma on the dev branch, so I need to recompile some Elixir when I update. This actually OOMed when I kept Prometheus + Postgres up at the same time (I'm also running Prometheus/Grafana there for monitoring). And it was only 80ct/month more on Hetzner Cloud to double my RAM :D I think an actually feasible thing to do would be to install a k3s or similar on the server and migrate it all to containers, then utilize something like gitlab.com CI to build container images of Akkoma so I can just run them without the need to compile on my infrastructure.

This whole open source & self-hosted stuff is kind of amazing. My tiny (2 Core, 4GB) server runs my social media (Akkoma), headscale controller for VPN, gogs for git, syncthing for file-sync. At home and at my parents place I have a Rasberry Pi for Home Assistant that is hooked into my headscale network, too.

Okay so this is now cozy and all, still don't know what to post though :D


I guess might as well share some of the fun I had last week:
- Garmin Fenix 7: Much nicer than the Apple Watch. For me. Ca 2w of battery life, amazing output regarding training ("after this workout and the bad sleep you had, don't train for the next 3d")
- Flatcar container Linux: Seems to be amazing and the future for the platform we build at work. Super tired of running mutable OSs
- Lupine lighting systems new front bike light for the roadbike: 10/10. You pay for it, but this is better than what I imagine.
- Roadbiking: Should have never made the break i made, feels so good.

Reading what @cure53 wrote about the last two years of your work feels kind of surreal, very good in this case though 👯

@bitboxer Oh, klingt gut, ja. Ich leite das mal an die Entwickler weiter :)

@luebken „leider“ baue ich die Cloud selber :D

@luebken ich vermute es wird erstmal ein Postgresql Cluster (Vmtl Crunchy operator) und ein Rook Ceph S3 bucket der damit verwoben wird (für backups). Da frag ich mich zB wie ich Connection secrets kriege und so

@luebken Danke! Sieht aus als müsste ich das einfach mal durchtesten, mit der Doku bin ich eher verwirrter ob das alles so geht.

@luebken Sach ma - Ich überlege grade Crossplane einzusetzen. Grobe Idee: Ich will ein paar Kubernetes-native Resourcen als Paket schnüren (also cloudfrei einfach normale Resourcen) - habe ich die Doku richtig verstanden dass ich mit so einer XRD machen kann?

@cedi i know that feeling very well, my last job drove me to record bike rides. I’m way less on the bike now, so: maybe an indicator to change something? :D