Sheogorath

Am I understanding you correctly that you want to have basically the equivalent of multiple docker networks attached to a container?
That's not how Kubernetes is set up. It's not impossible, Multus can do a lot here, but usually you just have your Pod Network and if you want to restrict traffic, you use firewall rules on top. But maybe the upcoming release with switching out k3s with docker-compose will change that once again

2024-08-15 20:01:36

- reply

I decided to provide some more details, and as we can see, it's mainly the postgresql database pushing the boundaries here. I'm sure some smart people can optimise on that :)
https://mastodon.s3.shivering-isles.com/mastodon/media_attachments/files/112/967/340/196/090/684/original/932b005d313e854f.png

2024-08-15 19:41:15

- reply

My single-user HA setup with oauth2-proxy, a patroni postgresql cluster and keydb cluster as redis replacement running 3 sidekiq instances, 2 web instances and 2 streaming instances, is at 2.67GiB without any optimisation.
So, I guess one can manage to get under a gigabyte with some optimisations.
https://mastodon.s3.shivering-isles.com/mastodon/media_attachments/files/112/967/250/705/413/548/original/cc86a51c6cccb3b8.png

2024-08-15 17:52:23

- reply

tailscale expects such a delegation for self-hosted SSO over OIDC.
And you can use it to redirect things like blog mentions, or use any account handle on a domain you own to redirect to your mastodon account on any mastodon instance.
Have the domain example.com and want @[email protected] to point to your gardening profile on gardening.example.net? Webfinger can do that for you.

2024-08-15 17:38:31

- reply

my hero! Wanted to build that for a long time, but always found something else to do :D

2024-07-28 23:06:47

- reply

This feels like these toolboxes with custom cut foam for the tools. I always wonder: How do you handle it, when you buy a new tool?

2024-07-26 11:00:13

- reply

a very lightweight one, yes. But it still uses all other Container formats and features.
Kata-container is a bit more straight forward to understand given it runs full VMs, while gvisor does a lot of smart things to keep things a bit lighter.
https://www.youtube.com/watch?v=pWyJahTWa4I

2024-07-25 23:39:13

Looking forward to migrate my Kubernetes setup to a micro-vm environment:
https://github.com/siderolabs/extensions/pull/434
Already started to run pods in gvisor and now moving to KVM-based gvisor runtime, will be even more fun.
#taloslinux #kubernetes #gvisor #kvm

2024-07-24 23:53:11

Decided to built an extra line of defence into my #immich setup: https://git.shivering-isles.com/shivering-isles/infrastructure-gitops/-/tree/8f9dc2880a5548db10dce32dfb697ba323674baa/apps/base/immich/monitoring-integrity
Now running sha256sum over all pictures on a daily basis to make sure there is no silent data corruption. With the exception of the past 14 days, to allow some cleaning up to take place without triggering an alert.
#selfhosting

2024-07-23 17:12:00

- reply

Wait until you see docker diff ;)

2024-07-23 02:50:17

- reply

I guess it has to do with "being privileged"? I can life just fine with the status quo and not worry about regulation to make my life harder.
But once they take on FLOSS, I can see how easily they screw that up. Just look at all the source available licenses out there trying to do that. And recent pushes by the government to provide more liability due to the "supply chain security".
It's a classic "better the devil you know than the devil you don't".

2024-07-23 02:38:27

- reply

As a company I would plead innocent on the compliance aspect, because they have no control over that.
You can of course argue that a lack of quality control on the update for the EDR was source of the issue and therefore the company is liable.
But I would argue with food safety standards as we have better cases for direct liability of managers there. And you can argue that an update is like a batch of bad food.
But again, I think this will just bankrupt the company.

2024-07-23 02:24:57

- reply

We do update codes, yes, bringing the people to justice? That really depends on the origin.
An Endpoint Security solution should never be able to kill people. If it does, the person/organisation in charge of putting it in place failed to provide reasonable safety mechanisms.
I do agree that warranty wavers are a problem here, but at the same time, I don't know if I want them to go away, given how essential they are to FOSS.

2024-07-23 02:12:39

- reply

To be fair, I just go through all of this with my rather limited legal knowledge about German laws, which are probably not really applied here, so I might just be straight up wrong.
Also there are a lot of things you can still throw at them, and there are enough parties to sue to just try every angle :D

2024-07-23 02:05:30

- reply

Unless we can prove that manager X said to push this update, I don't see how they would be directly liable. However, if the company goes bankrupt or doesn't manage to fix its value, they could be liable for mismanagement by their shareholders.
Regarding the auditors, they check what's in the standards and the standards are usually written by some form of government body. I think we will see some updates to some standards but not more.

2024-07-21 16:15:37

- reply

Higher is better right? So we win! Wait, what were we going for? ;)

2024-07-21 12:41:35

- reply

Danke für die Erinnerung, dass ich meinen Duolingo account löschen wollte :)

2024-07-20 22:00:30

- reply

Given that these are probably more load config files than actual software updates, and basically "anti-virus feeds", it's neither unusual to update them on Fridays. Usually you pay them for exactly that, so you don't come back from a nice weekend and your office PCs and airport Terminals became little bitcoin farms over the weekend.
It is a bit upsetting that they didn't do much/any testing on these, but it's also not super surprising.

2024-07-19 12:55:37

- reply

Sadly, "odd" is the wrong choice here. license is a well defined term and implies that something of value changes hands. The same way a donation is a defined term, where donations are usually bound to a goal/target and can't be used for other things.
To put it quite simple: I would argue that calling it a license is a big problem when it comes to consumer rights.

2024-07-15 19:18:22

- reply

kebab-case. Why? Because your container names are DNS names at the end of the day, and DNS doesn't know case-sensitivity.

2024-07-13 20:11:10

- reply

Naive Frage: Macht es überhaupt einen Unterschied ob "Superreiche" keine Steuern bezahlen, weil sie ihr Geld in Ausland schaffen gegenüber dass sie keine Steuern zahlen, weil keine erhoben werden?
Also aus gesellschaftlicher/wirtschaftlicher Sicht. Denn wenn nein, dann wäre Steuern erheben ja die bessere Wahl, weil dann zumindest die Chance besteht, dass was reinkommt.

2024-07-11 20:52:28

- reply

Mhm, interesting, I think there exist consistency issues. I get a lot of things double in my notifications right now 🤔

2024-07-11 20:27:57

- reply

The encryption part (2) is on me, I didn't read the migration guide properly: https://github.com/siderolabs/talos/blob/main/website/content/v1.7/advanced/migrating-from-kubeadm.md

2024-07-11 17:37:50

My goodness what a bumpy ride. Switched one of my nodes to #TalosLinux, considering to migrate to it, but hit a few bumps and a major roadblock.
1. longhorn doesn't like it when the ephemeral partition is wiped.
2. talos generates an encryption key for secrets in etcd but never tells you about it, which might break your cluster during migration.
3. while talos encrypts its install partitions, encryption for additional disks is not supported.
Now I'm contemplating how to continue…
#kubernetes

2024-07-10 17:25:01

- reply

This way:
https://git.shivering-isles.com/shivering-isles/infrastructure-gitops/-/tree/e6de9763dc760d1fd33848f735bdd0de86a83ad2/shared/components/keydb
It's a generic component that can be included in any kustomization. It installs a keydb helm chart, with all the basics :)

2024-07-09 00:16:48

Started to swap redis instances for #keydb. Especially for #mastodon since it has been a problem multiple times, that as soon as "the right #redis instance" dies, all mastodon-web pods become unready.
This should be fixed now in theory .-. lets see what reality says in the next few weeks.

2024-07-02 23:13:01

- reply

Well, when project/team management becomes a Birmingham screwdriver.

2024-06-26 03:53:01

- reply

Just added some more resources like old CNAME records, a link to archive org as well as a link to the GitHub thread discussing the payload.
Interesting part here: I haven't come across a second source for the payload yet. it always circles back to the GitHub issue from 5 days ago.
#polyfillio #polyfillioattack

2024-06-26 00:24:17

- reply

As soon as it's popped, yes. Before that, I don't think so.

2024-06-26 00:21:22

- reply

Honestly I'm curious how many websites this will break 👀
How common is an up-to-date ublock Origin with an outdated browser that relies on polyfills?
I'll keep an eye out for the disputes :D

2024-06-26 00:12:29

- reply

I started to collect some useful resources on my blog: https://shivering-isles.com/2024/06/helpful-resources-polyfill-supply-chain-attack
This will be a growing article regarding the polyfill.io supply chain attack.
#polyfillio #supplychain #infosec

2024-06-25 23:00:46

- reply

My first contribution upstream to get it into the ublock Origin default ruleset: https://github.com/uBlockOrigin/uAssets/pull/24255
No idea, where this will go, but hey at least it's a start.

2024-06-25 22:59:23

If you want to help a lot of people to avoid a supply chain attack, get cdn.polyfill.io blocked.
Write it into your uBlock Origin rules, block it in your DNS, do what you need to do.
https://sansec.io/research/polyfill-supply-chain-attack
And also watch your DNS traffic for funny domains 👀
#infosec #supplychain #security

2024-06-24 20:37:11

- reply

How much do you trust McDonalds to have the freshest and best quality food?
I think for a lot of people the situation with Google, Meta and Friends is the same. They are everywhere and you get the same quality everywhere, but to get the really good stuff, you go somewhere else. It's just that people understand/care less about computers than they care about food (rightfully) and so we end up with quasi-monopolies.

2024-06-24 19:57:57

- reply

Uhm, GDPR constitutes these in the EU. Of course, we can all talk about how well or unwell things are implemented there, but we have the authorities in place.
And only now do I notice, that you are in the US 😅

2024-06-24 18:52:10

- reply

We trust random people all day, every day to make our food, to handle our water, to make sure every days life works.
It's not as bad as it sounds.

2024-06-24 12:00:46

- reply

What time frame do you expect? Not that it's just your expectation vs the default time for renews :D
https://cert-manager.io/docs/usage/certificate/#renewal-reissuance

2024-06-24 11:40:49

- reply

Huh, what kind of issuer are you using?

2024-06-23 16:42:05

The linux world becomes an amazing place if you stop considering yourself entitled to anything and just enjoying all the work that people have put into this free software world. Everybody just doing their part, what they like, what they enjoy. And you can happen to join and share what you've done. Maybe people will like it and pick it up or maybe they don't and just ignore it.
It's fine, just enjoy your fork or go back to their version.

2024-06-23 16:25:30

- reply

ist dein Datum kaputt? 12024?

2024-06-23 16:20:57

- reply

Was mich natürlich auch frustriert und ich glaube hier könnten Medien besser arbeiten: es werden viel zu oft einfach Aussagen übernommen, statt sie einzuordnen.
Es werden Gesetze verschärft, aber wie viele Menschen würde das denn betreffen? Wo kommen die Forderungen her? Wer hat solche Forderungen zuvor schon mal gestellt?
Und vielleicht auch mal ein revisit machen. Ein Jahr nach der Verschärfung was ist rausgekommen? Hat es den gewünschten Effekt gehabt?

2024-06-23 16:08:44

- reply

Wie Robert Habeck es formuliert: "Es sind gesellschaftlich relevante Fragen."
In den Unterhaltung in meinem Umfeld hört man immer wieder "die Leute sind hier ja zu Gast" was den Anspruch an das Verhalten widerspiegelt. Und sobald es nun heißt, dass da Menschen straffällig werden, sollen sie nun ihr Recht Gast zu sein verlieren.
Dazu kommt das Misstrauen gegenüber unbekanntem das in Jedem unterschiedlich stark ausgeprägt ist.
Alles Spielerei, aber es beschäftigt Menschen.

2024-06-22 15:12:04

- reply

I mean, I'm running around a dozen database in my Kubernetes cluster at home. We have seen Zalando and other do it in the thousands. The reality is, that automating HA databases has proven to be easiest on Kubernetes for me.
But that doesn't mean it's the ultimate answer. There is this point, where you start to reduce automation in order to improve availability, since automation itself becomes the risk. If your database is at that point, K8s is probably a bad choice.

2024-06-22 15:05:10

- reply

Because you already run a well managed database cluster and running things in Kubernetes has no benefit?
Maybe because you want to keep your clusters ephemeral and stateless?
It's hard to argue that databases are not fit for Kubernetes given that Kubernetes can be used beyond Kubernetes patterns.
But you can of course easily find Kubernetes environments that are not fit for databases.

2024-06-22 01:33:28

- reply

yes I did, but I've selects their own difficulty here and with the upstream compose setup it should be easy mode. So no need to worry :)

2024-06-21 10:57:02

One could think that a positive outcome of all these AI companies scraping the web could be well curated archives of the internet.
But the reality is that it's all thrown on a pile behind locked doors that are expensive to open and undermine any well intended non commercial archiving efforts to preserve the web, since the commercial scraping for AI often doesn't care about the unofficial social contract between website owners and content collectors. As a result things are locked up. .-.

2024-06-20 20:22:29

- reply

Für alle die sich wundern as ich meine, die Burn rates sind mal kurz explodiert und es waren Dinge offline.
https://mastodon.s3.shivering-isles.com/mastodon/media_attachments/files/112/650/329/572/669/575/original/3e67a1e5cf9522bc.png

2024-06-20 20:18:03

Holpriges Calico update. Aber der Cluster schüttelt sich gerade selbstständig wieder zurecht.

2024-06-20 13:34:19

- reply

There is a mechanism to limit government power: Constitutions. Changing them is very complicated. It doesn't completely prevent it, but if you have to change the constitution to introduce a law, that's a big barrier.

2024-06-20 12:50:52

- reply

I'm already looking into better postgres client config as alternative :D

2024-06-20 12:42:24

- reply

My sidekiq instances were confused about the writability to their postgres target. I thought that the primary switched, but it seems like it didn't so I'm wondering now, how they got confused.

2024-06-20 12:18:22

Mhm, maybe I should set up pgbouncer for Mastodon.

2024-06-19 21:16:59

- reply

You could build all images from one repo. Docker supports giving it a git reference. E.g.:
docker build https://git.shivering-isles.com/sheogorath/hcloud-dynfw.git
Just with a bunch of image buildss in one
gitlab repository.

2024-06-19 20:03:46

- reply

Was that to renew or was that the initial price? I see it quite often that domains are cheap the first year and then explode on renewal .-.
These zone file entries are expensive .-.

2024-06-19 19:55:33

- reply

Well, you can just pass it on to your guests ;) Unless…

2024-06-18 17:54:43

- reply

> It's an explicit ideological attack on the whole idea of what Linux is and who Linux is for.
May I ask, because I honestly don't know: Who is Linux for?

2024-06-17 19:58:28

- reply

You can use kubectl wait for single objects if that's interesting for you. You can define a condition and wait for it.
https://kubernetes.io/docs/reference/kubectl/generated/kubectl_wait/

2024-06-17 18:59:59

- reply

No, I use Kubernetes foe my home infrastructure :D but I do use quadlet on Silverblue for containers running in the background of my desktop (like syncthing). That works fine :)
If you aren't into systemd, then looking into docker and docker-compose is probably the better time investment.

2024-06-17 18:32:59

- reply

if you are comfortable with systemd, maybe try podman with quadlets :) they bring containers close to the systemd experience.
Depending on your taste that might be a good thing :D
https://www.redhat.com/sysadmin/quadlet-podman

2024-06-17 18:08:16

- reply

let me help you a bit. I went with 6TB drives 2 years ago. It works fien and I still have enough space.
But there are basically 2 TB gone because I store the image of an old NAS on a drive and I added desktop backups that have up to 10 years of archiving using pika backup. Which also takes some space.
Then 6tb for jellyfin, 3tb for k8s service backups. It works out fine, there is still space until the 5 year mark is hit for the NAS.
https://gitops-docs.s3.shivering-isles.com/hardware/nas.html
Maybe 8tb drives are enough.

2024-06-17 17:36:46

- reply

buy the disks, the data comes by itself. Maybe you rip all your DVDs and built an own Netflix using Jellyfin. ;)

2024-06-17 17:06:02

- reply

Mit Hedgedoc kann man einfache Präsentation machen. Geht in der Regel ganz gut :)

2024-06-16 20:18:13

- reply

immich is probably the way to go. Not sure how it works performance-wise on a raspberry Pi. But it's probably fine :)

2024-06-16 15:58:39

- reply

For those wondering how to do this in Terraform with e.g. Hetzner:
https://git.shivering-isles.com/shivering-isles/infrastructure-gitops/-/blob/441b1072adaa305d648b066c1e42158e3dba53fc/terraform/hedgedoc-demo/firewall.tf#L22-34
There you go :)

2024-06-16 14:49:14

- reply

But what about the law of conservation of energy?

2024-06-16 12:14:44

- reply

Yes, something I also do and recommend is to isolate it network-wise from the home network.
My jellyfin gets its volumes from outside the container, can't talk to anything but DNS internally and the web to get plugins and metadata and all that. So unless one gets a container breakout, internal systems should be fine.
https://git.shivering-isles.com/shivering-isles/infrastructure-gitops/-/blob/968c44f00a1027a084076b958c1081aa82ad7894/apps/k8s01/jellyfin/kustomization.yaml

2024-06-16 12:02:10

- reply

so the categories of your risk profile are the same:
- password stuffing,
- unauthorised access for files
- arbitrary code execution due to a vulnerability
- …
What you do control with VPN vs Reverse Proxy is the amount of people able to execute these attacks. Any automated scanner vs. devices on the VPN (these can also be compromised).
What you want to do is decide the risk Jellyfin poses to you when it's successfully attacked and take steps to isolate jellyfin.

2024-06-16 11:51:10

- reply

Take your time, there is no urgency ^^
Currently there is only one instance and you would basically go ahead and deploy this multiple times as a result each instance would be individual.
Currently the way it's set up TLS wise TLS gets terminated on the loadbalancer and and then nginx does a reencrypt and to kitten with rhw localhost certificate. It's mainly because I didn't want to handle double issuing of certificates.
If I could just mount kitten a certificate it would be awesome.

2024-06-16 04:41:15

- reply

I think it's mostly done.
https://kitten.shivering-isles.com/
Image source: https://git.shivering-isles.com/shivering-isles/infrastructure-gitops/-/blob/2a0863b5d5b09c73d3fdbd550d81814634705fec/images/kitten/Earthfile
Deployment: https://git.shivering-isles.com/shivering-isles/infrastructure-gitops/-/tree/968c44f00a1027a084076b958c1081aa82ad7894/apps/k8s01/kitten
Having TLS being necessarily terminated in kitten doesn't really make it easier and I was quite confused that changing the would trigger a new secret to be created. (I noticed it's the same as origin security, but still was confusing to me at first.)
I think i if you manage to give each Pod a service of type load balancer, you can already run it like that.

2024-06-16 00:35:53

- reply

Yeah, as written I know why. Technically and from an organisation point. I participated in the change proposal a few versions back. :D

2024-06-16 00:32:33

why does visudo open nano on Fedora…
I mean I know why. But you know, WHY?!
If I type visudo, I expect to see vi, not nano :D
Certainly a violation of the Principle of least surprise!
#Fedora #linux

2024-06-16 00:28:01

- reply

It inspired me to look a bit closer into kitten. I decided that my weekend project will be to see if I can beat the 10 seconds startup time for kitten when using containers on a K8s cluster.
I know this is not your goal with small-web org, but for an organisation like a university it could make it trivial to just spin up hundreds of instances for all their users with a volume attached and run quite efficiently.

2024-06-15 20:11:06

- reply

how does one have half a wedding?
And if there are two half wedding is the still a marriage at the end?

2024-06-15 09:54:29

https://www.tagesschau.de/wirtschaft/ende-paydirekt-100.html
Den Einbruch an Transaktionen Anfang 2024 ist sehr einfach zu erklären: Man hat giropay auf paydirect umgestellt hat.
Dadurch wurden zum einen Details zum Kauf, also z.B. welche Artikel, an paydirect übertragen, als auch die Notwendigkeit erzwungen eine Paydirect Account zu benutzen.
Im Vergleich dazu: das ursprüngliche Giropay brauchte nur die Gesamtsumme des Einkaufs und man bekam eine vorausgefüllte Überweisung im Online-Banking der eigenen Bank.

2024-06-14 22:11:19

- reply

Well, now you have a reason to go outside and spend some time away from your computers.

2024-06-14 22:09:07

- reply

I read it's already resolved, but I think no one explained what went wrong.
kubectl config set-cluster
will not configure the cluster you will be using. It'll configure a cluster that can be used. your context selects what identity and cluster you use, so you want to check:
kubectl config current-context
and set the current context using:
kubectl config use-context <your-context>
otherwise configuring servers won't change anything.

2024-06-13 19:32:29

- reply

ich finde es super, dass wir solche Angebote bekommen, gerade für Menschen die deutsch lernen ist das eine ideale Quelle für Information und tägliches üben.
Aber Respekt an die Redaktion und die Sprecher. Dinge in einfache Sprache zu packen und vorzutragen ist nicht einfach und wirkt zumindest auf mich sehr anstrengend, weil es dem natürlichen Sprachfluss zuwider läuft.

2024-06-13 19:25:07

- reply

you don't necessarily need an SD card. rpiboot is a wonderful tool: https://shivering-isles.com/2024/05/screw-up-raspberry-pi
You can basically boot from your PC/notebook through the USB-C port that powers the Raspberry Pi.

2024-06-12 10:28:45

- reply

Tja, da hat unser Bildungsbürger leider den Artikel nicht gelesen. Aber keine Angst, am Ende ist er nochmal in einfacher Sprache zusammengefasst.
Denn einfache Sprache hat nichts mit einer Behinderung zu tun. Es gibt auch viele Menschen die auch einfach kein Sprachniveau C1/C2 haben, weil sie z.B. zugewandert sind oder noch deutsch lernen.
Diesen Menschen qualitativ hochwertige Nachrichten zur Verfügung zu stellen ist doch eine gute Sache, oder?

2024-06-12 00:59:29

- reply

I looked into it a bit, and there is a lot of ruby and postgres things going on. Apparently ruby creates the date in the local timezone, and postgres does all the timezone stuff internally only. So ruby has to tell postgres the timezone or it won't be aware of it.
Further it seems that at least for the API, there is also tinkering with the time of the returned date, resetting it to midnight.
It's quite mindbending :D It could explain the issue, but maybe it doesn't. I don't know.

2024-06-12 00:28:59

- reply

Timezone :)
The created_date is recorded in UTC and then just formatted in your local date format, which means your join date might shift if the time different to UTC is larger than the time you are into the new day.
https://github.com/mastodon/mastodon/blob/c93aacafdea188cac791b62a32e3117a7dc3e9cc/app/javascript/mastodon/features/account/components/header.jsx#L441-L444

2024-06-12 00:21:58

- reply

Well, that makes a sustainable business model. And funnily enough, I've happened to have webinars that were pre-recorded, and at the end you have different people pick up questions and answering them. Making it closer to a visit to the cinema, than an interactive thing. But surely quite profitable for the company making the webinars.

2024-06-11 17:35:56

- reply

naja, was heißt Nutzer verloren, eigentlich hast du Platz für nette Menschen geschenkt bekommen ;)

2024-06-11 03:40:58

- reply

It's more bootc-based: https://docs.fedoraproject.org/en-US/bootc/
From the user experience, it's close to coreos and Silverblue, but the developer experience is quite different.

2024-06-11 00:04:37

- reply

Productivity!

2024-06-10 11:05:57

- reply

1 Woche, 3 Tage, 2 Stunden und 34 Minuten.

2024-06-10 02:43:00

- reply

oh, that won't fly. You can make general statistics the typical efficiency, but most likely the selection of hardware and system configuration will have a much bigger impact on the energy usage than the software itself. Especially for server-side software.
But kde does some software efficiency assessments: https://eco.kde.org/blog/2024-04-12-sok24-energy-measurement-lab/

2024-06-09 16:50:48

- reply

Link to the project: https://git.shivering-isles.com/sheogorath/hcloud-dynfw
For a cloud provider of your choice you probably have to and want to use a different implementation all I'm saying it's possible.
You could just use a VPN or a jump host that uses some other authentication. You could also just not run SSHd at all, which I can also recommend.

2024-06-09 16:47:39

Oh funny, my hcloud-dynfw tool has birthday today. 3 years by now, that I use it and cut down my SSH log from basically all irrelevant login attempts.
No SSH exposed to the public internet except of GitLab's SSH port for git.
The concept is simple. You take have this CLI running in background, it dynamic reconfigures a firewall rule of Hetzner Cloud to your current public IPv4 and IPv6 domain. You use this firewall rule for SSH.
Simple, clean and trivial to implement. No VPN, no other stuff.

2024-06-09 16:30:14

- reply

while people on Hackernews lose their marbles over this, since it may or may no triggers bans for entire ISP networks, the right call is to use your firewalls to block the majority of trials for illegitimate access anyway.
Suddenly this becomes your second layer of defense and quite sensible because someone in a network with access to your SSH ports is trying something odd.
Practical implementation? https://git.shivering-isles.com/sheogorath/hcloud-dynfw

2024-06-09 15:38:23

- reply

Die EU ist übrigens auch dafür verantwortlich, dass immer mehr Hersteller ihre Geräte länger mit Software Updates versorgen. Hier gab es eine Regulierung, die Hersteller dazu verpflichtet Software Updates für die erwartete Lebensdauer von Geräten bereit zu stellen.
Eine "Right to repair" Regulierung ist ja gerade joch auf dem Weg.
Mit anderen Worten: Die EU spart euch Geld, wenn ihr nicht immer das Neuste haben wollt. Die EU macht sehe viel Regulierung im Interesse von Konsumenten.

2024-06-09 12:38:55

- reply

for vercel and netlifly, yes, GitHub pages, no. They already host so much stuff, serving a few websites won't make a difference.
It's actually a good chance it'll help to preserve a lot of pages beyond the author's original interest without becoming a financial or maintenance burden.

2024-06-09 12:36:18

Die EU ist dafür verantwortlich, dass du jeden Hannebambel fragen darfst, was er mit deinen Daten macht.
Sie ist auch verantwortlich dafür, dass du deine Arbeitszeiten dokumentieren musst. Klingt nervig, bedeutet aber du kriegst wirklich Geld oder Freizeit ausgleich für deine Überstunden.
Die EU ist auch dafür Verantwortlich, dass du jetzt rechtssicher "Memes" posten kannst. Also zumindest ein bisschen. Es hätte besser sein können, hätten mehr Leute besser gewählt!!

2024-06-08 23:04:23

- reply

For me the most important questions are:
- Will this become a huge maintenance burden? (e.g. does it introduce new dependencies?)
- Will it be useful to more than just the author?
- Does it work? Is there a demo of it deployed?
- And does it align with the general direction of the project?
And otherwise, just get towards deprecation management. If things break and no one steps up, be prepared to deprecate and drop features :)

2024-06-08 16:07:24

- reply

You underestimate a lot of wifi solution for cafe's and restaurants like this one:
https://purple.ai/blogs/cafe-social-wifi/
When it has a captive portal that requires you to accept terms of service, most likely you accept that your traffic is data mined.

2024-06-08 13:42:50

Gedanke: Wer gültig wählen geht entwertet damit die Stimme von allen anderen Wahlberechtigten ein kleines bisschen.
Wählen gehen ist also total egoistisch.

2024-06-08 13:30:49

- reply

Isn't octocat the official GitHub mascot?

2024-06-08 02:45:03

- reply

well that's a turn I didn't see 👀
I mean, it remains to see if they pull that on #immich after just announcing that it'll remain AGPL. Quite concerning.
We will see, I'm quite sure that when things go south we will see an immich fork.

2024-06-07 10:06:04

- reply

by the way, something I thought at the end: Something everyone can do immediately, check how much client traffic hits your http port on the API and start identifying who it is and why they hit it.
This way you can slowly fix it without necessarily breaking anything. You should get quite valuable insights there like User Agent, source IP and API endpoint they are talking to.
And you can just use something like goaccess and point it at your currently log files.