I think the problem with Linux desktop can be perfectly illustrated by how you create a shortcut.
On Windows, you right click the desktop, click new -> shortcut. A window appears with a paragraph explaining what you're doing and a single field to select the filepath. There's a button that reads "next." You click it to go to the second window, in which another paragraph tells you to pick a name for the shortcut. There's only one field, here, too. The next button became a finish button. This whole process just to fill 2 fields: filepath and name.
I'll take the Linux method any day -- it lets me put custom commands, including command line args. Could I just write a script and point my shortcut to that instead? Sure.
At the end of the day, making things idiot proof can be how some software developers go -- personally I can't stand it when it's done, however, as it makes actually accessing configuration options more bothersome than it frankly needs to be. Personally I can't stand MacOS, Windows is a time waster with a lot of how it does things, and most of the mainstream linux distros in the past few years (especially since the advent of systemd) have been moving scarily in the same direction.
Thank Bob for Slackware -- if Pat ever gives it up it may need to be a move to BSD for me.
I've never heard of DuckDB. I haven't used SQLite a lot, but one thing that I noticed it lacks, and this is pretty important, is that you can't add a foreign key constraint after a table is created, which is bad if you change your schema later. I don't even understand why this limitation exists.
One typical way to work around this limitation is to create a new table, copy the data there, and drop the old table. Copying from table to table is in general very fast.
Primary keys and foreign keys are rarely used in data science workloads. In our performance guide, we recommend users to avoid using primary keys unless they are absolutely necessary [1]. In workloads running in infrastructure components, keys and constraints are more important, and they are supported by DuckDB (but currently some of them can only be added when the table is created).
>if we want a true indie web, we must be prepared to fight for it. Hope is not enough.
This argument would be far more convincing if I saw some gifs of dancing bananas, rainbow colored text, a main text column that spanned the entire screen width, a repeated background image right under the text, a music playing widget that autoplayed music, and it changed my cursor to a pencil or something.
Right now, I'm looking at a medium post that's not on medium. This is what I see.
Is it really your personal website if it doesn't contain your mixtape?
Eventually it might. The same as some software is built with implicit linuxisms and systemd-isms that requires patches to build and run on BSD-like systems and other Unixes, there will be a point where developers will make assumptions on the presence of X and Wayland that'll make it difficult to make software run on X.
We already are at that stage with browsers - running Firefox on one machine while displaying the UI in another is a huge pain because the rendering is happening on the X client and it's pushing pixels across the network to the X server.
It is not across the network. A Unix domain socket is hardly something that could be called "across the network". Also I am pretty sure Firefox, for example, uses the MIT-SHM extension (Shared Memory).
>“I wish we were better at having a standardised desktop that goes across all the distributions. This is not a kernel issue; it’s more of a personal annoyance of how the fragmentation among different vendors has perhaps held the desktop back a bit,” said Torvalds in a TFiR interview in 2018.
This Torvalds guy is right. If you try to write a tutorial for how to do something in Linux, you will be in a world of pain, because users aren't using Linux, they're using apps that are using Linux.
Can I write "click on the start menu at the bottom left"? I can't. Because first of all nobody even officially calls it the start menu as Windows does. Sometimes the taskbar is at the top, other times it's at the left. Can I say press windows key (I mean, the super key (I mean, the meta key)) to bring up the application launcher? I can't do that either, because it depends on the distro's default configs. Can I say Ctrl+Alt+T brings up the terminal? Who knows! At least, surely, all of these distros are GNU/Linux so I can rely on bash commands being available everywhere, right? But then if I'm writing a tutorial for SSH'ing into a server, and the server is running FreeBSD, now the commands are the same but the arguments are different! And I know FreeBSD isn't Linux but it surely is weird that all these things that are similar but different are different from that thing that is also similar but different.
Meanwhile if you write a tutorial for Windows XP, it still works in Windows 11 because the system GUI and tools barely changed.
This is an advantage in volunteered end user documentation that Linux can't enjoy, because any step-by-step tutorial for Linux would have to be written for every single distro, or become or a ridiculously convoluted mess. And with Waylands that is only going to get worse.
Personally I'm not really counting on Linux desktop gaining popularity because, with all honesty, I found the GUIs of almost every Linux application I tried just terribly poor. Maybe this is because from I'm from XP era when lists and buttons were tightly packed into forms without all this auto-save switches nonsense, but in my mind, if Windows is losing users as they abandon their old GUIs and move into this same waste-of-space-fest (although that's probably not the reason for it), then maybe Linux should just make their GUIs like old windows instead of making them like new windows.
I don't see the reason why so many of these apps lack a main menu, for example. When an app has a main menu, and I can't find a function in the main menu, I just go "well, guess it just can't do that." But now I'm just wasting my time looking around the screen seeing if there is something I can click to do what I want. And right click often doesn't do anything. Where are my context menus? And where are my tooltips? This isn't designed for desktops, is it? It's designed for tablets. I just don't think going from Windows 11 to Windows 8 is much of an improvement. It's like you made a CLI tool without --help. In KDE Discover if I press F1 do I get a manual? No, I get the About screen.
Please do not make applications thinking the user can just Google it if they're stuck when if you Google "how to do X in Linux" you get 99 different solutions that don't work for your distro.
> Please do not make applications thinking the user can just Google it if they're stuck when if you Google "how to do X in Linux" you get 99 different solutions that don't work for your distro.
This is like searching "How to do X in Windows NT" and expecting to get results for Windows 11. Should be no great shock that when you use non-specific search terms the search engine returns non-specific results.
No, this is like searching "How to do X in Windows" because for a normal user it's Windows and Linux, not Windows and Ubuntu, Xubuntu, Kubuntu, Lubuntu, Linux Mint, Debian, Gentoo, Fedora, Arch Linux, Manjaro, etc.
Edit: I just realized Windows is so obviously Windows I didn't even write Windows XP, Windows Vista, Windows 7, Windows 8, Windows 10, Windows 11. I just wrote Windows. That's what not being fragmented means.
What I find odd is the human obsession with making AI as smart as a human.
In the early periods of AI image generation, you would see AI generate images so uncanny no human could make something like that. It was genuinely unique and otherhumanly, worth of being called art.
We "fixed" that bug so we could make the world's most expensive face swap app.
The headline made me think they were suing the tool, but they are really going after an AI offering made by Nvidia that used the copyrighted work.
>AI's rise has made Nvidia a favorite of investors.
>The Santa Clara, California-based chipmaker's stock price has risen almost 600% since the end of 2022, giving Nvidia a market value of nearly $2.2 trillion.
Me and Joe Nobody wouldn't be able to afford it, but I find it hard to believe a trillion-dollar company couldn't get proper rights to the data it uses to train its models.
The issue is whether or not you'd need to. I say no. Nothing was taken or copied. If you don't want your work seen, don't put it on the internet. They might have had a chance when Google started crawling the Internet, but that ship has sailed. Imo, crawling and making an index is more "stealy" than adjusting parameters.
>If you don't want your work seen, don't put it on the internet.
Reminder: anyone nerd enough to be on this site is on the top 5% of computer skill, and 20% of people in the planet can't even use a computer, and 30% that can wouldn't be able to perform a simple multi-filter task such as "find a sustainability-related document that was sent to you by John Smith in October last year."[1]
I believe if the web is meant to be open, it can not antagonize 50% of the people in the world who can't understand technology. It's just not in the spirit of the web to say "don't put it on the internet." The web was made to put things on the internet. When we can't do that anymore, the web is dead.
I can see the argument that if you train on a currently sold copyrighted work you have to buy a normal copy, you can't just download a torrent and still call it fair use (if you accept the concept of copyright at all...). So if someone could prove that Nvidia or whoever didn't own a copy (note who has the burden of proof) they could be liable for whatever an illegal downloading infringement is worth. Otherwise I agree that training on anything public is fair use.
Owning a copy of a book doesn't give me the right to make copies and sell them or even give them away. You can't even do that with publicly available material unless it has been put in the public domain or exceptions like fair use or satire apply. The model contains a copy of the work, and you can't distribute or publish your model without violating copyright. Offering access to this model as a service is legally in dispute, with AI companies generally arguing that liability for copyright infringement lies with the end user.
How is distributing a model containing copyright material any different to distributing a .zip file containing copyright material? Both contain an encoded, transformed copy of the work. And both can be used to reproduce the original work. The reproduction doesn't even need to be verbatim; just similar enough to convince a court that it is plagiarism.
Fair use generally only applies for parts of a work, and not using the entire thing. AI companies are claiming fair use, claiming that the processing is transformative. This needs to be determined in court. I think the court cases currently underway will show it is not transformative, because the original work can be recovered. Fair use is also determined on a case by case basis, so I think courts won't have the power to grant blanket permission. New laws would need to be passed for that.
> How is distributing a model containing copyright material any different to distributing a .zip file containing copyright material? Both contain an encoded, transformed copy of the work. And both can be used to reproduce the original work. The reproduction doesn't even need to be verbatim; just similar enough to convince a court that it is plagiarism.
Because of the substantial non-infringing uses. A court wouldn't throw away a model that can do non-infringing stuff 90% of the time if it can infringe copyright the other 10%.
> Fair use generally only applies for parts of a work, and not using the entire thing. AI companies are claiming fair use, claiming that the processing is transformative. This needs to be determined in court. I think the court cases currently underway will show it is not transformative, because the original work can be recovered. Fair use is also determined on a case by case basis, so I think courts won't have the power to grant blanket permission. New laws would need to be passed for that.
Fair use can apply to any possible allegation of infringement, not just parts of a work, as long as courts find the facts are in favor of it.
Whiile fair use is a case-by-case thing, courts can set precedent that will apply to all situations with similar fact patterns.
I think the problem with Linux desktop can be perfectly illustrated by how you create a shortcut.
On Windows, you right click the desktop, click new -> shortcut. A window appears with a paragraph explaining what you're doing and a single field to select the filepath. There's a button that reads "next." You click it to go to the second window, in which another paragraph tells you to pick a name for the shortcut. There's only one field, here, too. The next button became a finish button. This whole process just to fill 2 fields: filepath and name.
On Linux, you right click on the desktop, click "create launcher," and this appears https://i.stack.imgur.com/kFhFO.png