redjard

For UDP you’d also usually reencode packets as http3 warp ones […]

Is this similar to when you mask it as webrtc?

Sorry I probably meant QUIC not warp. It is related, I think webrtc now usually uses quic, so masking as quic or webrtc are probably synonymous, but honestly I don’t know enough about the details to say for sure.

Something other than DAITA?

I know people in Russia are reporting Mullvad gets fully blocked, so I don’t think DAITA is enough. As @chimp@lemmy.dbzer0.com mentioned, Hysteria is more in the right direction, though in more modern iterations/forks and together with another layer of obfuscation.
I don’t generally like to shout out names too much, since popularity among those that don’t need the tools makes them more cracked down upon for people that do need them. So I’ll just say that for the less obfuscated but more responsive route via rtc you use an updated hysteria as the outer layer, containing a fully homogeneous encrypted stream from a tool named after a reptile.

As per the GDPR or something else?

Mainly Net neutrality rules. Net neutrality just barely allows basic DPI, though it already heavily restricts what ISPs can actually do with that inspection. Wikipedia: EU net neutrality violations. Note that the 2007 case there happened well before regulations were strictened.

Size and timing of packets. Protocols generally have consistent patterns, so say always 6 packets with the last 3/4ths filled. Then some keepalive at regular intervals (i.e. tracker updates), establising new peer connections, …
Luckily this deep inspection is illegal for isps in the eu.

Russia, China, and Iran are using the patterns to effectively block all normal VPN and proxy protocols. There, people need special tools that send fake packets, randomize timing, and split and group packets into different sizes. All of course decreasing speed and increasing latency.
For UDP you’d also usually reencode packets as http3 warp ones so it looks like a video call.
Then the other end needs the same tooling to restore the normal packets.

Obfuscating it would be possible if necessary, but it’s better when you aren’t forced to maintain such an obfuscation stack.

4h at full speed. Uncapped is effectively an 8TB cap, so they delivered 0.6% of what they were supposed to.
My grandma has a plan like this and visiting her for less than a week detonates her (hidden) cap just from my phone backing up its apps.
My home internet sits at multiple TB a month despite backups etc. not even running through it since it syncs to my local nas on the same network.

On 50GB you coulnd’t even install a large game.

I’ve asked the author and they responded

[my question, translated]
Hi,
just a quick question about the 2nd latest blog-article “Five daemons or one binary” “Postfix and friends against Stalwart, tested side by side”:
[…] and it came into discussion if/in which way it was written/translated using AI/LLMs?

[response, german, translated by me]
Hi,

yes I use LLMs as support. But I test everything properly/carefully, [and then also(?)] by hand. I only trust the output after tests and when I understand it.

I released the tests for Stalwart today. https://www.behindtheconsole.com/2026/11/11/testing-stalwart-before-trusting-it-with-mail-the-test-catalogue/

Cheers, -mat-.

There’s no way that’s a german accent. There may be some human some ai parts, but that does still call into question if they made sure everything came out correct. Just a human translation from someone not familiar with every detail can already imply something wrong due to slight changes in wording if you are unlucky.

And the main original issue was “if they used ai for the writing, what else did they use it for?”. AI is a sign of low quality, and with lots of 100% slop creations posted on this com I wouldn’t read that sort of thing without someone vouching for the quality of the source (Which I take it you do?).

tldr Because of the ai slopocalypse I just don’t like being hit in the face by ai text from opening a random post.

PS: would a human let this graphical issue slip through?

I noticed that together with the presumed-ai text earlier, but chocked it up to mobile layouting. But it’s the exact same on desktop.

Can you give this a slop disclaimer at least?

ddg uses bing, and while they can filter text things to be surprisingly usable, they can’t do anything about the garbage image search.

Unfortunately we may be forced to remove all advertising from public spaces to make them safer to robots.

Use .0, it’s free real estate.

There are no entries in the certificate transparency logs for these intermediates in that period, indicating that CT-enforcing clients like Firefox, Chrome, Edge and Safari could not have been attacked.

I mean what do you wanna know?
What’s your concrete question, or how did you get confused by dbus?

Understand on what level?

The basic idea is that dbus is a type of communication layer and api.
Services can register there and programs see their registrations. dbus can in theory then launch a service too, so things don’t have to run all the time.
The idea is similar to having some known config file, that programs write to and that a service listens on or uses later. But dbus has more granularity.

It also often runs the other way, with system components presenting info in dbus and programs using it, to adapt to user configuration.

And let’s not even get into udev.

Point here is that dbus could be implemented via a command and an sqlite database for example. The command checks the database and constructs the dbus state from it.
But for likely performance reasons it isn’t, so this info is simply hidden in dbus, in ram, instead of being a file.

You can access dbus via a socket file ofc, but it’s not like you can mount it.

Even more confusingly, the internal ephemeral dbus structure with its apis and endpoints resembles a filesystem. In that way it is similar to the w!ndows registry, though not nearly as bad.

so yeah dbus is most definitely not files

edit: also there’s multiple dbusses, usually one per user and one for the system. Those two hold different apis and info. It’s just the same thing twice tho.

Mobile coincides with the idea of calculations and data being handled by 3rd party cloud servers instead of yourself.
That should be the time where the majority of performance shifted from being in peoples personal devices, to being in data centers owned by fewer large companies renting out access in various ways.

Not sure where else you wanna move the compute now for the next “innovation”. Even the bullshit about space datacenters isn’t really any different, it’s still compute owned by a company rented out.

nearly empty

half fuel

I’m sure a skilled politician could sell this as being basically the same ;)

Iirc it starts nearly empty and is only fueled up in the air, so basically no way around tanker planes.

Lin-ix

Loin-ughs

Check out dbus and notify.
And then unpack your dracut initrd.

It’s android version dependent. Only 14 and up support 3rd party providers.

Passkeys are supported on devices that run Android 9 (API level 28) or higher.

On many devices, Credential Manager stores passkeys to Google Password Manager by default. Users can choose other password managers as its passkey providers in the System Settings on Android 14 or higher.

Given the slowness of android version rollouts, this will be an issue for a long time.

I also think supporting older androids is pushing apps to do it the “wrong” way and making it google specific.

android doesn’t allow 3rd party apps to use passkeys nor autofill 2fa consistently. For passkeys, you are forced to use google services for it, or loose access, making it pointless. TOTP codes meanwhile can at least be copied and pasted manually from a password manager.