The world, preferably as a whole (as in Geneva convention-style), needs to firmly come down on one side or the other on whether deriving AI model weights from copyrighted material turns the weights, and more importantly, what is then generated from those weights, into derivative works of the training data--i.e., does the training data copyright apply to the products of inference?

Because, the current ambiguity seems incredibly unfair: organizations with powerful legal muscle can pressure smaller companies like Suno to adhere to their demands in the name of copyright law. Yet, more or less all large LLMs are trained on CC licensed text and open source software created by me and million others that legally require credit. If the music industry's copyright give leverage over Suno's AI weights, then *I* want to enforce my right to be credited on all LLM inference products that industry now uses. Or, alternatively, lets us all, once and for all, just agree that the derivation of AI weights are genuinely transformative so neither my nor the music industry's copyright applies here. Because, then at least I can use their stuff without being sued.

But the rule cannot be that it is the one with the largest legal muscle that case-by-case gets to decide this.

One reason this type of scam works well (though it is not specifically relevant to the Super Mario case since it is not yet released for Android at all) is the horrendous practice to completely hide apps not available in your region / compatible with your device. This makes any similarly named app show up as the "only option", and will easily fool people.

You know what, Google? If I am looking for an app, I rather you show me that it exists but cannot be installed, rather than have me dig through tons of search results in vain.

Many seemingly confused and/or contradictory summaries here...

Here is my attempt:
1. Company (Mojang) releases game (Minecraft) + multiplayer server under closed license
2. Third party uses decompilation to create derivative work of server. Creates GPL project (CraftBukkit) in violation of both Mojang's license and the GPL (all code must be GPL). Mojang does *not* try to stop the unlawful redistribution.
3. An individual (Wolfe) contributes GPL code to this project.
4. Mojang hires devs of CraftBukkit and thus gets 'some kind of control' over CraftBukkit.
5. Wolfe submits DCMA takedown to those who redistribute CraftBukkit becasue it is in violation of his code's GPL, since it is distributed integrated with non-GPL code.

Note that *redistribution* is the key for the GPL license. The important question is thus: did Mojang ever *redistribute* CraftBukkit themselves? If they did, at that time they released the *decompilated* server code at whatever license it was specified. But, if there is nothing resembling an official Mojang release of a Bukkit binary or code, then nothing of this matters for them. The only people who then are in the wrong are the ones who redistributed a combination of a GPL and Mojangs closed source. They are liable both to Mojang (who are not currently complaining) AND the people who's GPLd code are used (one individual who now IS complaining). Per the DMCA takedowns they have to stop this redistribution.

Furthermore, IF Mojang did distribute this binary or code themselves, it appears some people (possibly Wolfe?) argues that this requires them to also open the original source code of the game server. That is an actually very interesting non-obvious license discussion. Personally, I don't *think* it has legal standing. The fact that Bukkit was built on top of the decompiled code makes Bukkit more of a normal derivative work from the decompiled code than the original minecraft server code. It would be a widely different matter if someone contributed obfuscated code to an already existing GPL project and then redistributed the combination.

Plot twist: The frog says: "I would have said: I have no idea what you are talking about". Explanation: the puzzle was created today. The every second day rule only applies starting from the day the puzzle is created.

Actually, asking "if tomorrow I ask you" is equally unsafe, since it could be that this is the final day before the puzzle is dismantled. To be safe you need to go with something more creative like "true or not, if we pretend this puzzle is still active tomorrow, and I were to ask you ..., etc."

But perhaps a more elegant solution is just "if the days you are lying and telling the truth had been reversed, and I asked you ...., etc."

This makes no sense! If you know the Frog always lies, just ask him "Where are the Riches and Power?" and be on your merry way through the opposite door.

Shouldn't the phrasing rather be something like "You can ask the frog one question, but every second day it lies, and every second day it tells the truth."

Now it is possible, but non-trivial, to formulate a question that tells you which door to use. (But not if the Frog is dead, though.)

I suppose the plot twist may be that both doors open to the same room, and the puzzle was designed by someone wanting to teach you the lesson that riches and power leads to certain destruction :)

Gnome3/gnome-shell now allows for a huge amount of customization with plugins, which really aren't that difficult to develop. Whatever "vision" you have about how the desktop should behave, why not simply implement it as a Gnome3 plugin?

With carefully selected, already available, plugins, you can more or less turn Gnome3 into the look and feel of Gnome 2, if that is what you really want; so I really do not understand what all the "Gnome 3 is so bad compared to Gnome 2" comes from. It seems all that is needed to solve that problem is for someone to prepare and maintain an easily installable "plugin cocktail".

The only major "desktop problem" I currently see is the duplication of effort between Unity and Gnome 3. I hope someone re-implements the Unity panel + dash as a Gnome 3. I would love to run Gnome 3 with dash.

I can see the point others often bring up about OnLive being "impossible" based on the total cost of bandwidth. But how can the light speed latency make OnLive impossible? Whatever theoretical calculations you can come up with are trumped by me actually playing Batman Arkham Asylum on Onlive and not being disturbed by the lag.

I think the pilot episode did motivate Eli's place on the ship, and he *could* have been a great character if the series had stayed consistent with the picture of him as a "highly unmotivated genius". Previous SG scientists have all been highly motivated and hard working, driven by a need to prove themselves. Eli could have been a nice contrast to this, with someone who is brilliant but really doesn't care much.

But you are right, in the rest of the episodes he was turned into just "a slacker teenager" who doesn't show any of the genius that brought him onboard in the first place. The writers pretty much use him as a button-mashing monkey for times when the story keeps Rush away from the console.

I am with you all the way on that the ship and its role in the mythos was tragically underused. Had they focused more on that, and tied the plot to it, the series could have been brilliant.

You make a good point on the crew-crew interaction. However, it may still be that this was caused by the "main plot" being so thin in most episodes, as it did put more focus on the soap opera part. I mean, also in SG-1 the character interactions could be cheesy at times, but this was less evident as the focus typically was on an engaging plot. A counter-example is one of my least favorite SG-1 episodes 'Solitudes' where O'Neill and Carter spends the whole episode stranded in a cave, with O'Neill dying. It is one of the few 'crew against nature' plots in SG-1, and to me, it mostly falls flat.

My take at some reasons for failure:

1. Most of their viewers identify with Eli (the slacker nerd genius), but he ended as a minor support character, often just tangentially involved in the plot. He should have been SGU's Rodney.
2. Unlike previous Stargate iterations and BSG they tried to pull off 'crew against nature' plot lines rather than 'crew against enemies' . To get such plots feel like 'action' is really hard. A lot of them (especially in the beginning) was "crew lands on planet, somehow gets stuck, must get back in time before the ship leaves". There is only so many times you can do that before it becomes repetitive.
3. Point '2' got even worse since the planets often were ridiculously uninspired, "Desert planet", "Freezing planet", "Jungle planet", etc.

All this said, I think the show was heading in an interesting direction. I'm sad to see it go.

Thanks for the tip. I didn't know about this (needs a plugin for thunderbird)

But I couldn't find a standard for redirected email, is there a rfc for this? Essentially it is like Ccc: + setting a custom from address. To me this really proves how this Aprils fools joke really isn't a joke. Rather, it would be nice with a standard for headers that covers this functionality.

You could always use "Forward", which includes the original message along with the list of original recipients.

Which usually requires you to add an awkward "Hi, I forgot to send this to you" to not make the inline headers too confusing. Sure, this is what I do today, but it is less convenient than Ccc: would be, and exposes my mistake, which I'd rather avoid if I could.

Sounds more like you'd want to recall the original, and re-send the revision.

I am absolutely opposed to a feature that would allow people to alter email I have already received. I often use my email as an historic record of events, and would hate if I could not be sure that it is immutable.

Also, I'm pretty sure there is no standard for doing this across mail systems. To roll it out would require a major revamp of email as we know it today, since it requires some careful form of cross-realm authentication. On the other hand, the Ccc: header is just a straightforward feature to implement in the MUA.

With a CCC, the original group still doesn't know Alice is invited.

True, but for me this usually is not as important as to point out to Alice that she is not the only one getting the email. In the rare occasions when it is critically important that everyone is aware of everyone else, a followup email with "I forgot Alice" would be motivated.

Personally, I'd like support for multiple Dcc: headers: Disjoint Carbon Copies. I want to send the same message to multiple groups of addresses where I want those in one set to know they were all copied but want to hide that it was sent to the other group, and vice versa.

If the Ccc header was implemented, you could easily do this by sending the mail multiple times. Send the mail once to just the To:/Cc: recipients. Then put the To:/Cc: recipients in the Ccc: header, and for each set of Dcc: recipients send a copy with them in the Cc: field.

Since I can't figure out a way to emulate Ccc: with Dcc:, my vote is with Ccc:.