You are right my last sentence was an approximation. The gain of H3 over H2 is theoretical not practical.
895 karma · joined October 30, 2021
Here "quicker" was about CPU load and well defined.
Please re-read too.
The only field that is computed in UDP is checksum and the same exists in TCP and it must be recomputed each time someone actually re-route the packet (eg: bridge to VM) since TTL is decreased.
So I doubt your assertion.
_____
Writing my comment I understood what your are talking about. There is a bunch of encryption done at user mode in HTTP/3 that doesn't need to be done in user mode. In HTTP/2 it was sometime done in kernel mode (kTTL), so was quicker. The slowness comes for the CPU needed it to be copied out of kernel mode. I didn't follow the whole story so I trust you on this.
> HTTP/2 is fully encrypted, so you need all your application servers to have a key and certificate
Nope. h2c is a thing and is official. But the article is right, the value HTTP/2 provides isn't for LAN, so HTTP 1.1 or HTTP/2 it doesn't matter much.
HTTP/3 however, is fully encrypted. h3c doesn't exists. So yeah, HTTP3 slower you connection, it isn't suited for LAN and should not be used.
BUT if you actually want to encrypt even in you LAN, use HTTP/3, not HTTP/2 encrypted. You will have a small but not negligible gain from 0-RTT.
It does. Just use `--http2` or `--http2-prior-knowledge`, curl deduce the clear or not clear by `http` or `https` URL protocol prefix (clear the default).
Reference: https://limitesmaritimes.gouv.fr/thematiques/espaces-maritim...
And there is not much military, it's more about scientists here than military people (about one hundred permanent people, space/climate scientists and biologist). The island is full of basalt yes, like most of the island, and the island is BIG. But if you are gonna nuck a french island, you should ask the French.
And it's the more important French island of the French Southern and Antarctic Lands (after Adélie Land if you count it as an island).
Btw, we should be able to make the bomb ourself.
Zero additional anthropic warming result in progressive global decarbonizing thanks to carbon sink. (If the carbon sinks still exist when we reach 0 net emission and the permafrost hasn't thawed yet)
For reference, Germany was the third biggest market for Tesla (1/10 of China). France was 4th (1/10 of China too) and Canada was 5th (1/11 of China).
It's not the end of Tesla, but it's politically related.
https://www.bloomberg.com/news/articles/2025-02-03/tesla-sal...
https://roadgenius.com/wp-content/uploads/2024/09/tesla-mark...
It's not even a discussion for us. That is fact, you or the ADL (I don't care who they are) can say anything they want, the fact that he did it and can't ignore the meaning of this make him a Nazi. Adding to the context of supporting Neo-nazi political groups and the ideas that he is pushing, there is absolutely zero doubt why he made a nazi salute.
https://www.theguardian.com/technology/2025/jan/21/the-gestu...
Tesla sales in Germany will drop again in 2025, and it's politically related.
By useful I mean, to the world. That it affect the world in a good way. Maybe it's not the best technology to do something but replace the best way in an CO2 effective way. Maybe it is not a clean technology but increase overall fairness.
I don't know, but it had to have a good income to the world in a way.
No predicted productivity boom (check last US data), no GDP boost yet (again last data). Even LLM enthusiast like McKinsey or Goldmansachs expect nothing before 2027.
And it's not about LLM, it's about the whole AI progress. That is, obviously, a revolution.
https://www.goldmansachs.com/insights/articles/ai-may-start-...
But just to be clear. I'm denying something said to be obvious. I should not be the one who give sources about something that doesn't exist. If there is a productivity boom, I may not have seen it. Show it to me.
And the only way to assert that they are is to get numbers (big or small) and to compare them to alternatives.
LLM are polyvalent. But in most of the tasks they are not the most efficient way to do the task.
Want to play chess ? Use Stockfish or Leela. Want to do image recognition ? SAM or TinyViT like models. Want to know if your are sick ? Go to the doctor or at least do a search on the web.
Yes, there is tasks where LLM are perfect for (speech analysis/classification for example). But omnipotent chatbot isn't one for example.
If there were a revolutionary use, we would have a productivity boom. We don't. This article is from 2021: https://www.technologyreview.com/2021/06/10/1026008/the-comi...
It's not about any individual usage. It's the global technology that is yet to prove to be useful and that already have bad for the environment.
Any new usage should be free of impact on the environment.
(Note: The technology of LLM itself is not an environmental disaster, but how it is put in use currently isn't the way).
- Half of the European inhabitant will live in a country with far-right head of state by the end of 2025.
- China will not become the first economy of the world but will invade Taiwan.
- Bitcoin will fall (and ETH/BTC will raise to at least 0.1).
- One Piece will not end in 2025 but in the next few months.
- USA will see another CEO shooting.
- India will become third largest economy in the world.
- China will outperform all other competitors in AI.
- Open Source model to solve simple math problems will be accessible (by the end of 2025).
It's very good. Some points that can be improved:
- Search inside the description of the bookmark, it doesn't. - Update to a new version of hoarder. Since the software isn't stable, it's a real problem. - Related to the previous point => More archive formats.
Otherwise, it's a very good software. Easy to use, nice front-end, good UX.
A real prediction will be the impeachment of Macron. It may happen if he fall short of candidate to be prime minister.
It's again a French speaking initiative (but from a French-Switzerland research collaboration) about creating a recommendation algorithm with a lot of good properties and that scale (check the research paper on arxiv.org : 2107.07334).
It's also a plugin to help then have more data and to be useful for the users (by having an other recommendation algorithm on youtube), a win-win collaboration.
No, that not what I mean. What you can't check is that they push something to someone specifically to make that person react in a certain way. That is not verifiable. Regardless of whether the content is true or false (that doesn't matter actually, or should I say it's less important whether it's true or not, obviously fake news are a problem but less important than that).
> Adding new information to your life will modify your view of the world.
Yes. And you should be able to choose what you add to your life, what you watch freely. Not a recommended algorithm that nobody understand how it works.
> You are just using a feature of content to argue it should be banned.
I'm sorry ? I don't understand what you are saying here. I read it many times, but I just don't understand what you mean.
> But content is not opium
It's debatable. Nearly 100% of TikTok traffic is recommended (it's an estimation) and as or some years ago(2015 ?), about 70% of Youtube was recommended (given by youtube). That mean people do not expressly choose what they watch and particularly, most of the traffic is compulsive. But it's debatable, you are right "content is not opium" but sometime it's consumed as if it were so how to deal with it ?
> and free speech is sacrosanct
Once again, yes. It's not about the content that is on TikTok, sorry if didn't make it clear. It's not about free speech. You can still say what ever you want. It's again too much power of TikTok.
> I am fundamentally against anything whose goal is to avoid "mass manipulation" that is essentially saying we should ban things that get popular which we don't like
That absolutely NOT my point. Effectively, my point promote to be against "mass manipulation" but absolutely not "we should ban things that get popular which we don't like". Remember that I also regret the ban. I think it was justified but I don't think it's a solution.
As a French, I learn early in life « Je ne suis pas d’accord avec ce que vous dites, mais je me battrai jusqu’au bout pour que vous puissiez le dire. ». Which can be translated "I don't agree with what you say, but I'll fight for you to be able to say it.". We both agree on that and as I learned, our definitions of free speech are slightly different but the differences are not relevant here I think.
For example and to finish, a "not so good solution" but something that could have been done to not ban TT is to switch the recommendation algorithm to the one of (the old) Twitter (or the current one of Mastodon), where people could have followers and tweets were shown on a timeline from they followers.
Now, I will again take twitter as example. The fact that Twitter (the new one, X) have an Open-source recommendation algorithm make it not ban-able for example.
Not really, twitter/the-algorithm isn't up-to-date and the models must be shared along with the source code but it's an example. What I say is twitter doesn't have a lot to do to make it any ban unjustified, If someone say "there is too much power from the platform (twitter) and there is manipulation from it", Twitter just have to update the source code (and the models) and it's verifiable (with hard work, but not impossible). That the algorithm is biased is not the point, you have the right to consult media you know the algorithm is biased. The point is that the algorithm must be verifiable.
Thank you for the debate. If I don't respond quickly enough, please send me an e-mail at "ache-hn at ache.one".