treffer

1. Everything must be visible. A diff between the release tarball and tag should be unacceptable. It was hidden from the eyes to begin with.

2. Build systems should be simple and obvious. Potentially not even code. The inclusion was well hidden.

3. This was caught through runtime inspection. It should be possible to halt any Linux system at runtime, load debug symbols and map _everything_ back to the source code. If something can't map back then regard it as a potentially malicious blackbox.

There has been a strong focus and joint effort to make distributions reproducible. What we haven't managed though is prove that the project compromises only of freshly compiled content. Sorta like a build time / runtime "libre" proof.

This should exist for good debugging anyway.

It wouldn't hinder source code based backdoors or malicious vulnerable code. But it would detect a backdoor like this one.

Just an initial thought though, and probably hard to do, but not impossibly hard, especially for a default server environment.

Ubuntu still ships 5.4.5 on 24.03 (atm).

I did a quick diff of the source (.orig file from packages.ubuntu.com) and the content mostly matched the 5.4.5 github tag except for Changelog and some translation files. It does match the tarball content, though.

So for 5.4.5 the tagged release and download on github differ.

It does change format strings, e.g.

   +#: src/xz/args.c:735
   +#, fuzzy
   +#| msgid "%s: With --format=raw, --suffix=.SUF is required unless writing to stdout"
   +msgid "With --format=raw, --suffix=.SUF is required unless writing to stdout"
   +msgstr "%s: amb --format=raw, --suffix=.SUF és necessari si no s'escriu a la sortida estàndard"

There is no second argument to that printf for example. I think there is at least a format string injection in the older tarballs.

[Edit] formatting

“Say you changed out fluorescents to LED light bulbs, that’s a capital improvement. You’re not replacing lightbulbs, you’re enhancing”. That's the basis for the claim that lightbulb changes are CapEx.

This is an improvement though. A rack is usually 10-20kW @ equinix, that's well within reach for redoing larger areas of lighting. They also bill for overuse. Great, they improved the building to make more money.

The other claimed scandal on that front is that they wouldn't change one light bulb at a time, but rather turn it into a larger project and then do it. This doesn't sound like a bad idea either. Avoid tiny maintenances, rather re

I wouldn't be surprised if their battery claims have a similar underlying issue, freeing up space in a similar way that lights free up power.

I don't see how this shouldn't classify as an improvement or proper management. And it's a good example of the quality of this report.

This one works in private mode. I am using Firefox on mobile and it worked that way.

There is also a plug in available on gitlab (iirc) that can auto-circumvent it on most sites. But I find that questionable and I try to keep my plug-in list minimal (it's a horrible supply chain attack vector).

I would give it a 50:50 chance.

The other option is that the scraper got lucky with a tiny glitch (whatsoever).

On the one hand I bet github does everything to keep stuff secure, on the other hand I can't believe there wasn't a single glitch in the last few years.

And if there was a glitch then regular automated scrapers are a pretty likely siphon.

Heck I should check if any of my private repository show up.....

I was surprised too. Then I checked my ryzen laptop and .... yep, max speed 6GHz.

I haven't noticed before either.

But given that my laptop pulls this off at a way lower TDP makes me wonder how competitive that Intel chip is, especially as the articles title already says it has a huge power consumption.....

Full enumeration is useful for security research. For blocking purposes it can sometimes be helpful to enumerate it.

And on your question: it is most often the case that a subdomain is owned by the same entity, but it is not a given. The most common thing is a CNAME to some site of a vendor. But you can go all the way to full subdomain delegation (this is the same as going from com. to facebook.com.).

Regulatory worried?

The EU stated they will investigate this. And somehow Apple decides less than a week later that the feature will return.

Looks like regulators should not be worried about Apple complying in the medium term.

DMA has a catch-all: any circumvention can lead to penalties. You can try to play games like this, but it will just lead to fines.

I would expect a few more U-turns, especially on the fee structure.

Just look it up.

Here is the pure number of facilities in Germany (2021)... Verdi (the relevant union) is collecting this data and distributing it.

17 fulfillment centers (logistics) 47 distribution centers (logistics) 7 sorting centers (logistics) 2 amazon fresh centers(logistics) 2 customer service centers 3 AWS centers 4 development centers

19k of the 28k workers are working in the fulfillment centers.

As others have pointed out: logistics.

E.g. the German postal service employs some 600k people. DHL is another 600k. The logistics union has like 1.3M members. That's like 2% of all employees in Germany.

Those are labor intensive jobs. With minimal training.

It's a huge sector that you usually don't see (well, except for all that truck traffic).

Virtual private server == anything that is not a whole machine, usually referring to small VMs.

In hetzner speak you want a "shared vCPU server".

That's roughly what others offer as VPS, just more clearly spelled out (IMHO). You get fractional CPU usage that should be burstable and roundabout one core.

Just a few points....

1. Some decisions are better made than delayed. Most choices are reversible. So get over those being be important/relevant.

2. Everything else _should_ be thought through. Push for it. If that's not possible then you have to live with the consequences. These are most often sales traps though, see e.g. "scarcity" in "Influence: The psychology of persuasion" which covers deadlines. If you fail for these then please read the book or listen to the audio book.

3. Winning an argument is not just about thinking about arguments. See "The art of being right" for a set of arguing tactics that will make you look like you are dumb.

4. Fast good action is often a matter of turning it into a routine. That's what you do for incidents (gamedays), fire drills etc. If it matters then better be prepared!

But on top of everything.... I've been doing on-call for years. I can tell you that bad decisions under time pressure and emotional stress are often bad and far from optimal. It's ok. It's human.

The thing that makes me a bit uneasy is that it looks like you are stressing yourself on top of everything. You can get professional help for that too.

This should IMHO summarized in the related git commits.

Yes they become paragraph long. But that's the point. That's where I search for "why". And not just "why", "when, why and with what reasoning?".

Imagine if there were some interference that could derive a comment based on history and current code!

Go to the cloud, it's cheap. And when it's not you just turn off parts of your system.

I am somewhat suspicious that that's not really the solution.

These court orders might also happen if someone _claims_ that crimes have happened.

The legal system has to come to that conclusion, which requires an investigation.

Yeah it must be an important claim. That's of course completely unacceptable and illegal. But it is one way such warrants could happen.

So I got curious, nice questions.

RFC6838: Expected uses for the "application" type name include but are not limited to file transfer, spreadsheets, presentations, scheduling data, and languages for "active" (computational) material.

"File transfer" sounds to me like the distinction is: if you would expect to show it in a browser directly use audio/text/image/.... - otherwise it's application.

This sounds OK for yaml. You _can_ still serve it as text/plain if you want to show it as text, but as yaml it's probably for download purposes.

Btw, the RFC discussions are open. You can join an IETF mailinglist and start discussing such topics :-)

I can't even find "cost" on that page. Only one rather tiny side note that you could get past the free tier quota.

I don't think that's a proper warning on costs.

The complaint says there should be a warning that processing fees can be high. Go to the front page and check out the links. Nothing really about cost. Someone follows that path and 14k gone without a word about it. That's the path that people are sent down from the website. It explicitly talks about using BQ for analysis.

A simple "running queries over the whole dataset can cause significant costs due to the size of the dataset" should be enough. And I think that's a valid and fair point.

The whole part of accusing Google should just be ignored.

Last time I checked the GCP prices vs. hetzner servers there was a factor of 10 in monthly cost. So I cant say this is a general rule of thumb. It is true though that hetzner can provide superb performance per $. Potentially below what you can do (unless you buy racks of hardware).

Also keep in mind that hetzner is mostly a beowulf cluster. See e.g. this press picture from their newsroom page: https://cdn.hetzner.com/assets/Uploads/IMG-0546-91.jpg

They do have dell servers, too. But don't expect to run on one of these nodes unless the machine type has Dell / PowerEdge in their name. This means the management capabilities are minimal (but usually enough). A fully licensed management card can do way more.

Their renting pricing is ridiculously cheap though. Hardware failure is more common than on cloud providers (see cluster hardware), support has been helpful in the past. Overall IMHO a good choice if you are price sensitive. But even then consider development vs. deployment vs. running (server) costs. Servers might make up less of the total cost of a service than you expect.

I never got this TBH.

In Germany and Europe it's simple: the price on the item is what you pay at the cashier, at least of you pick the cheapest payment option (cash).

Most don't bother to charge for other payment methods, a surprise price basically never happens.

Credit card payments are capped at 0.3% of the transaction value.

Large credit card cashbacks don't exist for that particular reason. But it's dead simple to know the price.

NVIDIAs margin is someone's money. I wouldn't say they don't understand it. I would say they need a good enough competition to get the margin down.

E.g. FB saying they want to buy 350k H100. That's just a whopping $14B price tag. With a >85% profit margin. While a fab is $20B.

Trillion? Sounds like anchoring to me. Nvidia has a market cap of $1.7T. You could literally buy NVIDIA for that. I read that as "a billion won't cut it, we need quite a few billions".

But it's not unreasonable that those hyperscalers throw in a few billion each.

Usually it's horrible business not to be best (see Intel/AMD). Because the margins are at the top. In this case though they want a whole range of products to go down in margin. Even a slightly worse chip might be worth it if it comes at a significant cost reduction. Especially if the optimal design is known!

In a sense the whole thing can fail at reaching the top or making lots of money and still succeed in bringing total cost down, potentially by 50% or more.

OpenPOWER happened a few years after Apple finished the transition to Intel.

The whole architecture is niche since then.

Linus Torvalds has a very interesting POV why x86 won. "Develop at home" issues. You are going to deploy to a system that is similar to what you built on. If you run x86 then you'll deploy to x86. And he points to that as the reason for x86 servers.

Home is today a x86 or arm computer (arm if you like Apple), perhaps some SBCs (usually arm, perhaps some mips), and some IOT (often esps, so xtensa / risc-v) plus some router/wifi device (arm).

RISC-V is scaling up on that axis. It is killing other ecosystems for embedded/iot. It's becoming useful for SBCs and low end desktop boards are on the horizon.

That's the scaling path that works. You are $20 away from trying it out. And it can scale all the way to an affordable desktop soon (Milk-V).

It's IMHO not "a lot of energy on RISC-V", it's a quickly growing user base. OpenPOWER lost that.

I wouldn't call it kernel based. It's not like this is an in-kernel emulation. I would stick with hardware virtualization.

KVM is a userspace API, and kvm-intel/kvm-amd are the drivers for the hardware.

You will be using hardware features. That's also why it is in the kernel: nothing but the kernel should have full unlimited access to the CPU to set this up.

So you could say it must be in kernel to keep the kernel secure. And the performance benefit is "just" exposed hardware features.

The kernel does not provide additional things. As far as I understand: you set up a dedicated memory space and handle traps that halt the execution e.g. when the VM talks to the PCI bus. (It's been a while since I looked this up)

But you need the pieces, especially virtual PCI devices. That's where qemu or VirtualBox enter the scene (or minimalist systems like firecracker). They provide a repository of virtual hardware and all the auxiliary methods to boot a virtual machines. You also need to emulate something like a BIOS or UEFI.

You can think of it as your CPU removing the need to emulate the very same CPU (and a memory controller). You still need to emulate the rest though! But running on the same CPU removes most performance penalties. You run at native speed.

Newer generations can even nest this. Having virtual machines in virtual machines. That's mostly useful for cloud environments so that the cloud provider can run kvm based VMs and you are still able to run VMs inside that VM.

I am somehow puzzled on how this goes into a downtime as implied in the article.

The control plane can be fully down - as in: you can shut it down - and everything continues to run. I've been in that situation multiple times with large clusters. E.g. one etcd node having disk issues, pretty much turning it into a lame service (even worse than true down). Kubelets got randomly regarded as non-healthy due to update latency. But everything continued to run. Another time API servers were leaking memory, and with that crashing, causing some herding that would crash each server as it comes up. No issues whatsoever, migrate to larger instances.

It's a pretty cool feature of kubernetes.

I am wondering what was done to let this cascade like this. The only thing I could imagine is that someone _wiped_ the etcd state, then brought it up, casing all things to go down.

It becomes more funny once you shift online. Basically every checkout has a "coupon" field.

Now you look for coupons, and some coupon aggregator of doubtful quality pops up.

You find something that works. Or give up. Done.

The coupon sites are usually of doubtful quality because .... Basically all affiliate networks forbid it. It's also the reason why "last click" attribution is horribly broken.

Funnily enough it's now mostly magazins & newspapers that run these coupon sites in Germany.

Or you make the case that the OS should ship a DNS server. It can handle DNS forwarding and iterative queries anyway (iterative queries as you often need them for dnssec) and the logic for serving DNS responses is ok'ish (DNS name dedup and length restrictions being the biggest complexity issue).

I am not super happy with systemd-resolved but it solves this particular issues. No requirement to use libc, but same (os configurable) behavior for all users.

Well, the good news is that there aren't enough of these a*** to destroy it.

When Wikipedia started and even like 10-15 years ago that wasn't a common belief. Especially in Germany. "There are some assholes on earth, how could you dare to believe something voluntary and open to everyone will succeed?!?" (others already posted that it's not unfounded given real world experience)

Wikipedia though is my preferred counter example for that.

It's not that. It's basically the old question: can algorithms format cartels?

I am strongly on the side of: yes, they can.

In a time where most offerings are public you just need to make sure you stay slightly above the p50 and prices will go up.

Similarly on the buying side of e.g. Ads most algorithms and platforms have a "bidding war avoidance". Stop bidding more even of it would make you money, in order to keep the proce low.

In all those cases you try to sync and rise /lower the price (as seller/buyer).

It's not about just estimating a price. It is doing that on a large scale with an algorithmic outcome of upwards price fixation.

The problem is this... The money cow is datacenters, and especially top of the line products where there is no competition.

Then fastest single core and multi core x86 cpus that money can buy will go to databases and similar vertically scaled systems.

That's where you can put up the most extreme margins. It's "winner takes all the margins". Beeing somewhat competitive, but mostly a bit worse is the worst business position. Also...

I put money on AMD when they were taking over the crown.

Thank you for this wakeup call, I'll watch closely if Intel can deliver on this and take it back. I'll habe to adjust accordingly.

Well, "seize" = court order or a police action.

If police shows up to seize something then you can either cooperate (hand it over voluntarily) or risk additional penalties (fines, prison time, ...).

I read this story as "law enforcement was able to figure out that they have bitcoins, so they knocked on their door, and the criminals opted to hand it over instead of prison time".

So yes, they transferred it, but I wouldn't call this voluntarily. It's just that the applied force is as visible as the seized bitcoins.