HN user

zinodaur

501 karma
Posts1
Comments158
View on HN

The query planners have diverged between MySQL and MariaDB; which one are you talking about? Both have received improvements in recent years, so the specific version is relevant as well. And at least both systems have supported index hints

MySQL - we use the index hints 100% of the time because the query planner (still) doesn't work

The real-world relevance is minor, especially since companies using external write-through caches are already making much worse inherent trade-offs in that area

I disagree - everyone knows that the cache value can be stale, and to be wary of it if your use case requires correctness. In my experience people tend to expect correct, up to date data from their database - which is why they usually have to opt in to read-replicas

Do you have some novel academic solution for this that you can propose?

MySQL doesn't support synchronous replication - so semi sync is the next best thing. Then you wrestle with it to make it not crazy (e.g. - only count a tx as acked if it was acked by a replica that has all the other txes). Its not rocket science - maybe the big fanfare about synchronous replication made sense when a disk write was incredibly slow, but adding 2 ms of network latency and 5ms to persist it on another machine in order to not lose user data seems like a pretty straightforward choice to me. And if write performance matters that much to you - you shouldn't be using a dinosaur like MySQL anyways

- Its query planner is ancient and broken - it comes up with very bad plans. Every time we rely on it, we regret it eventually

- you have to use locks to prevent mangled write transactions, instead of the db handling it. Jepsen report is scathing

- its pretty hard to set it up in semi-sync replication mode, a.k.a. "only return from a transaction commit when the transaction is present on at least one other replica". Once you've set it up, you have to deal with insane things like transactions being visible on replicas before leaders, or even worse transactions visible on replicas briefly but never visible on leaders or other replicas

- every knob and feature contains subtle bugs, sometimes ones as scary as index corruption, that you only discover after it has caused you pain

I don't know why you would choose it unless you were already running it

Any government, organization, or individual can abuse any technology. But you haven't made the case why opposing technology itself would prevent that, versus holding those individuals accountable directly

I think we should hold the individuals accountable directly. But we can't. The system is skewing further and further from the point where we could. Look at the Epstein files - everyone on the planet knows that there is a mountain of evidence condemning someone rich and powerful, and nothing will be done about it.

In the meantime, I want to stop handing weapons to the powerful people that we can't hold to account. I don't think we should stop all R&D - but I think "machines that act as if they can think for most practical purposes" are uniquely dangerous. I also used to think the "AI" companies were full of shit, until my work handed me a bottomless anthropic API key to use for claude code. They can successfully navigate novel situations using tools to interact with the world. Tasks like "find me 20 puritanical whitehouse staffers who are cheating on their spouses, using credit card / location history" are now costly only in terms of api tokens. Going the other direction - "Find the organizers of this protest. Using all the information collected by big tech, find an unrelated criminal offence they have committed".

AI doesn't exist. It is a marketing term used by grifters to sell their snake oil.

They've claimed the term, this is not a useful objection to make at this point. And everyone was fine with calling our shitty little computer vision handwriting parsers "AI algorithms" before LLMs.

We've somehow been able to do this for nuclear weapons which can literally obliterate civilization at the press of a button

Knowing what you know about nuclear weapons, if you ran into the Manhattan Project scientists, would you still be cheering them on? "Thanks guys, our democracies are so stable these will literally never be used for a nuclear holocaust, and they might have useful mining applications!"

Can you not think of any exceptionally nasty things the US government could do with the "machines that act as if they can think for most practical purposes"? Do you think maybe it might be a good idea to develop that technology after you have made sure that the government serves the peoples interest?

why start with Sam Altman?

Well Zuck has that big scary hedge, and I’m sure people have been going after him for ages.

I despise Mark Zuckerberg, but I don’t want to firebomb his house: I want his company neutered and/or broken up, I want him stripped of his ill-gotten wealth, and ideally I want him to face criminal prosecution and incarceration.

Great! Is the plan to wait until after the billionaires have their AI controlled military drone swarms to have this revolution? Because they already control your government - I don’t think you will achieve anything like this through legal means

Sibling comment already said it, but yes I was specifically alluding to Altman's decision to allow the US government to use their AI to choose bombing targets without a human in the loop - perhaps this is why the US government double-tapped[1] a school killing 160 girls, all younger than 12, when the school was clearly marked on google maps.

I also vigorously dislike the industry, but your stance 'I'm on the skeptic side of "AI"' is something you need to address - saying this in the friendliest way possible, you are wrong.

AI needs to be opposed, because the billionaires are going to use it to turn the world into shit, but if the best the AI opposition can muster is "AI isn't useful", we are fucked. It's extremely powerful and can do bizzaro things when you rig it up with tools - the kinds of things we need to prevent companies like Google from doing with it, no one is paying attention to.

[1] double-tapped: a phrase referring to the practice of firing a second missile after the first to kill any rescuers or surviving schoolgirls

Since 1985, Hezbollah has killed approximately 600 Israelis (if you count IDF soldiers during the occupation of Beirut). Israel has killed 5x that number of civilians in the last two weeks, if you count Lebanon as well as Iran. If you count soldiers...

If Canada and Mexico started letting Iran launch bombing sorties against US cities from within their borders, would the US consider them neutral?

2 Million a ship seems like a pretty cheap price to pay for the damage the us and Israel have inflicted on Iran - they cannot be made to pay it though, so I suppose the rest of us will have to (through marginally higher oil prices in the long term - much less than the spectacularly high oil prices the US war will cause in the short term)

I think what we should have learned from this is that it's extremely hard to "make a lesson out of" Iran if you depend on moving oil past their borders... the gulf states are much more exposed to this than the US is, and much less powerful.

They are also not neutral - they have been paying in to the US protection racket, and are discovering that their payments haven't bought much.

The whole world is hurting pretty bad right now when it comes to tech jobs - America seems to be hurting the least.

I'm not saying the tech job situation in America isn't bad - but the world dances to America's fiddle, and its frustrating hearing Americans complain about how hard their situation is while their boot is firmly planted on my neck

You're assuming things I didn't state. I don't particularly want mass AI surveillance at all

That's fair, sorry for that.

considering how much more dangerous a government's mass spying is to its own citizens living in it 24/7, it's not unreasonable for that to be the focus

The US government is actively trying to influence politics in my country and spending huge amounts of money to do it. The US government is a much larger threat to us than our own government.

All of our tech is owned and operated by US companies, which means the US government has read/write access to all of our data. If we attempt to incentivize domestic software production (e.g. by taxing imported software, or by stipulating where our data can be stored and who can access it), the US government will destroy our economy. This has played out a few times recently.

I can't believe we were so foolish as to let this situation grow. Its going to be a painful few decades.

Sibling comment summed it up pretty well; my country is considered an ally of yours, but even left leaning Americans seem to take it for granted that we deserve mass AI surveillance/blackmail/manipulation if there’s a chance it could benefit us citizens in the short term. I suppose we deserve it for being complicit in American crimes for so long

AI makes you boring 5 months ago

Using AI to write your code doesn't mean you have to let your code suck, or not think about the problem domain.

I review all the code Claude writes and I don't accept it unless I'm happy with it. My coworkers review it too, so there is real social pressure to make sure it doesn't suck. I still make all the important decisions (IO, consistency, style) - the difference is I can try it out 5 different ways and pick whichever one I like best, rather than spending hours on my first thought, realizing I should have done it differently once I can see the finished product, but shipping it anyways because the tickets must flow.

The vibe coding stuff still seems pretty niche to me though - AI is still too dumb to vibe code anything that has consequences, unless you can cheat with a massive externally defined test suite, or an oracle you know is correct

I wonder if we can trust Gemini to do its job well here? Whoever is being protected in those files obviously has the power to compel governments to do what they want - if Gemini started being a threat, I bet it would get some "alignment" help. Certainly its findings would be reported, as well as the identity of whoever was doing the prompting

Eh, if I'm paying someone to host my git webui, and they are as shitty about it as github has been recently, I'd rather pay someone else to host it or go back to hosting it myself. It is not absolutely required, but it's a differentiating feature I'm happy to pay for

All the same financial rails apply to crypto - enforcement is just lagging a bit.

E.g., you could do what World of Warcraft does - Gold can be earned/exchanged in game, and can also interact with the real world in nebulous ways. Using the hyper advanced technologies of relational databases and ignoring financial legislation, they have enabled ultra-high-throughput microtransactions, with the added benefit of not spraying the public ledger on to the desk of every law enforcement agency on the planet.

We should ban dynamic feeds that aren't based on explicit user action. E.g., Youtube should only be able to show search results based on search term, not search context. The recommendations should only be videos from channels you have subscribed to.

The dangers of algorithmic content are so obvious, and the only way to stop companies from doing this stuff is to legislate against it

There’s a chance that the current situation will start to resolve itself in 3 years and we go back to normal, however that might look.

I don't think it can - dependence on US digital infrastructure grew at a time where American stability was taken as ground truth.

How can an EU leader sit across the negotiating table from a country that can delete (if not read/alter) all of their data, and a willingness to exercise that access?

Even if Trumpism goes away, to know for a certainty that Americans won't do it again one election cycle seems like it will take a long time to establish.