The ubuntu:24.04 Docker image is only 77.30 MiB.

alpine:3.19.0 is 7.38 MiB.

Of course those sizes are without a kernel. Typical everything-included distro kernels are generally a few hundred MiB as they include drivers for everything that might be needed, but a custom build for known hardware can reduce that to just a few MiB.

Most of this is personal opinion and snobbery that I can’t do much about except maybe ask that you examine how anarcho-capitalist your takes sound.

Objectivist, perhaps. They’re the ones who obsess over controlling and monetizing free external benefits. There is no copyright in anarcho-capitalism (including “moral rights” etc.) so the GP doesn’t sound at all anarcho-capitalist while arguing for infringement of others’ real property rights to prop up their own artificial (non-rivalrous) “intellectual property” rights.

They didn’t say it shouldn’t have been developed. Improving the AI models so they can deal with this kind of malicious interference gracefully is a good thing.

The EULA also prohibits using Nightshade “for any commercial purpose”, so arguably if you make money from your art—in any way—you’re not allowed to use Nightshade to “poison” it.

You’re restricting speech whether or not you confine your censorship to only AI-generated images.

Correction: Fortunately, not unfortunately. A rule like that would prohibit any form of public / street photography, news videos, surveillance videos, family photos with random strangers in the background… it’s not reasonable at all.

Since you don’t understand, quotes denote emphasis or specificity, not emotion.

Actually quotes denote quotations. When used casually around an individual word or short phrase they generally indicate that the writer is emphasizing that these are someone else’s words, and that the writer would have chosen a different description. As in: These people are described as “teens” but are probably not only/mostly teenagers. That may not be what you meant, but it’s how that text will be read.

If you just want emphasis you might consider using bold or italics rather than quotes.

I’d settle for just the limits, personally.

The part that makes me the most paranoid is the outbound data. They set every VM up with a 5 Gbps symmetric link, which is cool and all, but then you get charged based on how much data you send. When everything’s working properly that’s not an issue as the data size is predictable, but if something goes wrong you could end up with a huge bill before you even find out about the problem. My solution, for my own peace of mind, was to configure traffic shaping inside the VM to throttle the uplink to a more manageable speed and then set alarms which will automatically shut down the instance after observing sustained high traffic, either short-term or long-term. That’s still reliant on correct configuration, however, and consumes a decent chunk of the free-tier alarms. I’d prefer to be able to set hard spending limits for specific services like CPU time and network traffic and not have to worry about accidentally running up a bill.

You mean “3. Object Code Incorporating Material from Library Header Files.”? That section 3? I think they’re using a bit more than just header files. Section 4 “Combined Works” is the one that applies here.

Also even if section 3 did apply they’d need to follow 3.b as well as 3.a and include the full text of both the GPL and the LGPL.

Technically it can be statically linked, but then you would need to provide artifacts (for example, object files for the non-LGPL modules) enabling the end user to “recombine or relink” the program with a modified version of the LGPL code.

Dynamic linking is usually simpler, though. And the DRM issues apply either way.

Section 6 of the GPLv3, which the LGPLv3 includes by reference as one of the required distribution terms in paragraph 4.d.0:

Convey the Minimal Corresponding Source under the terms of this License, and the Corresponding Application Code in a form suitable for, and under terms that permit, the user to recombine or relink the Application with a modified version of the Linked Version to produce a modified Combined Work, in the manner specified by section 6 of the GNU GPL for conveying Corresponding Source.

(emphasis added) There is the alternative of following 4.d.1 instead, but that’s only if the application links against a shared library already present on the user’s computer system—it couldn’t be distributed with the program.

GPLv3 section six offers five alternative methods of satisfying the obligation to provide source code. The first (6.a) applies only to physical distribution and must include source code with the physical media. The second (6.b) also requires physical distribution plus a written offer to provide the source code to anyone possessing the object code. The third (6.c) is the one I mentioned that applies only “occasionally and noncommercially” for those who received a written offer themselves under the previous clause. The fourth option (6.d) allows for the source to be provided through a network server:

If the place to copy the object code is a network server, the Corresponding Source may be on a different server (operated by you or a third party) that supports equivalent copying facilities, provided you maintain clear directions next to the object code saying where to find the Corresponding Source. Regardless of what server hosts the Corresponding Source, you remain obligated to ensure that it is available for as long as needed to satisfy these requirements.

The fifth and final alternative (6.e) pertains to object code provided through P2P distribution, with the same requirements as the fourth method for the source code.

The GPL in most cases only requires that derivative work must also be shipped with the same license. The source code from providers doesn’t have to be distributed by unity, it has to be distributed by the provider.

This is incorrect. The distributor of derivative works in binary form is responsible for providing the source code. They can refer to a server operated by a third party, but if that third party stops providing the source code the distributor remains obligated to ensure that it is still available. The only exception is for binaries which were originally received with a written offer of source code, where the offer can be passed on as-is, but that only applies for “occasional and non-commercial” distribution which wouldn’t work here.

Sure, they don’t rule the world. They only have the power to ban you (either the company per se or its individual owners, officers, and/or employees) from ever again doing any business in the EU. Which naturally includes business with any individuals or companies either based in the EU (as a seller or a buyer) or wanting to do business in the EU. Or from traveling to the EU, whether for business or personal reasons. Little things like that. Nothing too inconvenient. (/s)

They haven’t taken things quite that far—yet. But they could. It’s dangerous to assume that you can ignore them without consequences just because your company doesn’t currently depend on revenue from EU customers. The world is more interconnected than that, and the consequences may not be limited to your company.

Geoblocking in such cases would not be sufficient. For one thing your geo-IP database will never be perfectly accurate, even without considering that “data subjects who are in the Union” can connect to your site via proxies or VPNs with non-EU IP addresses. For another you still need to respond to GDPR requests e.g. to remove data collected on a data subject currently residing in the EU, even if the data was collected while they were outside the EU, and you can’t do that if you’re blocking their access to the site. For a newspaper in particular the same would apply to any EU data subject they happened to report on, whether they had previously visited the site or not.

They never should have made opt-in an option in the first place. All the legitimate reasons to store data are already permitted without asking permission (required for the site to function, or storing data the user specifically asked the site to store such as settings). All that’s left is things no one would reasonably choose to consent to if they fully understood the question, so they should have just legislated that the answer is always “no”. That plus a bit more skepticism about what sites really “need” to perform their function properly. (As that function is understood by the user—advertising is not a primary function of most sites, or desired by their users, so “needed for advertising to work” does not make a cookie “functional” in nature. Likewise for “we need this ad revenue to offer the site for free”; you could use that line to justify any kind of monetization of private user data.)

So you’re not remapping the source ports to be unique? There’s no mechanism to avoid collisions when multiple clients use the same source port? Full Cone NAT implies that you have to remember the mapping (potentially indefinitely—if you ever reassign a given external IP:port combination to a different internal IP or port after it’s been used you’re not implementing Full Cone NAT), but not that the internal and external ports need to be identical. It would generally only be used when you have a large enough pool of external IP addresses available to assign a unique external IP:port for every internal IP:port. Which usually implies a unique external IP for each internal IP, as you can’t restrict the number of unique ports used by each client. This is why most routers only implement Symmetric NAT.

(If you do have sufficient external IPs the Linux kernel can do Full Cone NAT by translating only the IP addresses and not the ports, via SNAT/DNAT prefix mapping. The part it lacks, for very practical reasons, is support for attempting to create permanent unique mappings from a larger number of unconstrained internal IP:port combinations to a smaller number of external ones.)

Not every work produced by a LLM should count as a derivative work—just the ones that embody unique, identifiable creative elements from specific work(s) in the training set. We don’t consider every work produced by a human to be a derivative work of everything they were trained on; work produced by (a human using) an AI should be no different.

Examples of local commands I might run in tmux could include anything long-running which is started from the command line. A virtual machine (qemu), perhaps, or a video encode (ffmpeg). Then if I need to log out or restart my GUI session for any reason—or something goes wrong with the session manager—it won’t take the long-running process with it. While the same could be done with nohup or systemd-run, using tmux allows me to interact with the process after it’s started.

I also have systems which are accessed both locally and remotely, so sometimes (not often) I’ll start a program on a local terminal through tmux so I can later interact with it through SSH without resorting to x11vnc.

Not the GP but I also use tmux (or screen in a pinch) for almost any SSH session, if only as insurance against dropped connections. I occasionally use it for local terminals if there is a chance I might want a command to outlive the current graphical session or migrate to SSH later.

Occasionally it’s nice to be able to control the session from the command line, e.g. splitting a window from a script. I’ve also noticed that wrapping a program in tmux can avoid slowdowns when a command generates a lot of output, depending on the terminal emulator. Some emulators will try to render every update even if it means blocking the output from the program for the GUI to catch up, rather than just updating the state of the terminal in memory and rendering the latest version.

Historically speaking, people have gone to the trouble of manually digitizing hard copy books to distribute freely. There were digital copies of print books available online (if you knew where to look) before e-books were officially available for sale in any form. That includes mass-market novels as well as items of interest to historians. Ergo, your scepticism seems entirely unjustified.

OCR is far from perfect (though editing OCR output is generally faster than retyping), but even without it we have the storage and bandwidth these days to distribute full books as stacks of images if needed, without converting them to text. The same way people distribute scans of comics/manga.