I have a 160TB minio cluster running for 4+ years who had dealt beautifully with node outages, one drive failure and the occassional hiccups on the datacenter.
I was okay with not having support because I am not part of their customer base. I was okay with not having the webUI, though I wish they made an option where the webUI would be available for some basic-tier paid customers. But I can not be okay with this move. They are just giving the finger to all the community. They never tried to work out a solution that could let smaller users to contribute or support.
I will seriously have to consider moving to Hetzner object storage.
Right now, my problem is that I can not update my minio cluster because I do not know of any trustworthy docker image that I can use, and the version I am on is exposed to (at least) one known CVE.
Read https://www.joelonsoftware.com/2006/12/09/simplicity/ and ask yourself if you are truly solving anyone's problem or if you are just looking for a way to rationalize the amount of time you are spending on a hobby.
The complete opposite. It's OP that's trying to "optimize the web server for reverse proxying and static file serving", when what we have out there is more than enough.
> or you're wasting time
"Wasting time" is not a problem. If OP is doing working on things because it brings them pleasure and they are hoping to learn from it, more power for them. What bugs me about these types of posts is when people are set on the "build a better mouse trap" mentality and want others to validate them.
It may sound "harsh" to you, but if I came up asking for "any type of feedback" when I'm trying to figure out if the idea is worth persuing, I'd be pretty upset if I kept chasing an invisible dragon because the community was more concerned about "hurting my feelings" instead of being upfront and give some warning like this might be interesting to you but it's not solving any real pain point. Keep that in mind when deciding if work on this will be worthwhile.
> It's OP that's trying to "optimize the web server for reverse proxying and static file serving", when what we have out there is more than enough.
I have optimized it, so it would be faster than the original server I have been working on.
> (...) give some warning like this might be interesting to you but it's not solving any real pain point. Keep that in mind when deciding if work on this will be worthwhile.
If you feel the project isn't solving a real pain point for you, you don't have to use it! I was showcasing my web server to interested people on Hacker News.
> It's good to have as many web servers as possible out there.
The problem space of "web servers to serve static files and reverse proxy" is fairly small, how many differing solutions and designs would be required to satisfy your idea of "as many as possible"?
At what cost? For what benefit?
Again: if OP wants to work on this because they take joy in it, fine. But be honest about it (to themselves and to others) instead of coming up with all sorts of ratioinalizations and biased comparisons when talking about the alternatives.
It isn't [strokes moustache] film per se but it is a movie. It's a moving picture with named characters and three acts. Believe me when I say that I'm a filthy hipster that would be first in line to take a dump on this flick but I kind of liked it for what it was.
Eh, I guess I am talking to militant anti-SUV people.
Allow me to rephrase:
- Your environment imposes restrictions upon you
- Even if you can control your actions, optimal choice is to move within those restrictions
- Doing things that attempt to move outside those restrictions are not optimal
- Some people choose the optimal path
- Some people are upset that the optimal path is chosen
Good grief, why am I bothering with nonsense so early?
> Doing things that attempt to move outside those restrictions are not optimal. Some people choose the optimal path.
Optimal for what?
> Some people are upset that the optimal path is chosen.
Person A chooses the "optimal path" (according to whatever definition of "optimal" A has) for their benefit. Their "optimal path" puts person B at risk and forces them to deal with unwanted costs and changes their environment. Do you think that person B is wrong to be upset about the choice of person A or not?
<< Do you think that person B is wrong to be upset about the choice of person A or not?
Oh boy. I am not responsible for you. By this tirade, you only demonstrate to me you are willing to make suboptimal choices so that you can feel better about yourself. That is cool, but don't drag me down with you.
By your logic, each time you breathe out CO2, it forces me to deal with unwanted costs and a change to my environment. Can you hear how ridiculous that argument is at its core?
Can you hear yourself and realize how ridiculous your reduction ad absurdum is?
Let me help you: taking your analogy to the other extreme, and it seems like you shouldn't be mad at anyone if they decide to light up a cigarette in an elevator.
I am not mad. At best, I am disappointed as I let the someone go by themselves as I don't get on the elevator. For every choice, a consequence. It is absurd that you think your response was a reduction at all.. Honestly, if you are on my side, please stop. You are explicitly not helping.
> disappointed as I let the someone go by themselves as I don't get on the elevator.
Ok, so you think that people are expected to just step down and be quiet about it. Others would certainly complain and rightly so.
Also, while you might feel okay about taking another elevator, we can not tell people "if don't like your pedestrian-hostile and accident-prone environment just go move away, or stop being a pedestrian".
I say "people are justified about being upset, because they end up facing the consequences and bearing the risks of the choices made by others" and you somehow imply am I saying this is an argument about "forcing" anything?
That is a seriously bizarre conversation. Peace out.
People in the EU are still using Instagram/Facebook/WhatsApp. Zuckerberg did a "ok, if you don't want us to track you, you can pay 12€/month" and everyone just smashed the "I consent to get my data mined forever" button.
Not to mention that we *still* have lobbying for chat control.
Every measure from the EU is, as always, meant to look like our beloved bureaucrats are doing something but absolute ineffective at changing the status quo.
Everywhere there is a choice. If, for instance, you don't want Facebook to have your data, don't use Facebook and block their domains using an adblocker to prevent leakage via cookies on sites that have their JS running.
The folly is that EU regulators have duped people into believing that they can read and even participate in free-as-in-beer social media sites without any of 'their data' ever being held by anyone else. And that the information of what a person did on a single site on a given day is "personal data" even if it isn't ever tied to your identity or ever used to do anything but try to show that browser a more relevant ad.
The root cause of both the general public's discomfort and the overcollection of data is the ad-supported model itself. The EU punted on that hard problem and instead wastes everyone's time trying to pretend the ad-supported models can function with 100% personal data control.
Things elsewhere are bad, but the EU is worse because it lies to people about the efficacy of its regulations and the whole apparatus only exist to make lawmakers and lobbyists a justification for their existence.
Let's stop pretending that the EU has done anything more than political theater.
> What do you suggest instead?
Break apart any company that has more than 150 employees (by employee, also count individuals working more than 50% of the time to the same company): https://news.ycombinator.com/item?id=31317641
The one thing I don't understand is this assumption that demand for GPUs for training is going to keep growing at the rate they grew so far.
I get the demand for new applications, which require inference, but nowadays with so many good (if not close to SOTA) models available for free and the ability to run them on consumer hardware (apple M4 or AMD Max APUs), is there any demand for applications that justify a crazy amount of investment in GPUs?
Inference will be cheapest when run in a shared cloud environment, simply due to the LLMs roofline. Thus, most B2B use cases are likely to be datacenter based, like AWS today.
Of course, cern is still going to use their FPGA hyper-optimized for their specific trigger model for the LHC, and apple is gojng to use a specialized low power ASIC running a quantized model for hello Siri, but I meant the majority usecase.
I do not buy this premise. I think it will end up being cheaper to simply run the LLMs directly on the user device.
I think that there are plenty of competitors in the "LLMs with open weights" space to essentially make the models a commodity, so all that is left is the compute cost and there is no way that someone will be running a datacenter in a way that is cheaper than "the computer that I already have running on my desk".
I nake your point every time this comes up[1] but its absolutely surprising how few business people, most of whom have some credibility in the form of qualifications or experience, actually recognise a value chain when they see it.
Apologies for the second reply, but it also occurs to me that reinforcement learning is the new battleground. Look at the changes between o1, o3 and GPT-5 thinking. Sonnet 3.7, Sonnet 4, and Sonnet 4.5. And so forth.
I expect models will get larger again once everyone is doing their inference on B200s, but the RL training budget is where the insatiable appetite sits right now.
I do not believe for a second that any of those people investing tens of billions of dollars are doing it to "get to AGI". They would only be able to profit from a AGI if it could be simultaneously (a) weaponized and (b) strictly controlled by one party, and there is no one crazy enough that these could be achieved.
If you tell me that people are pouring all that money into data centers because they believe that most applications will use some form of LLM or VLM as the main driver of machine-to-machine and machine-to-person interface, I'd be more inclined to buy it. But then I'd respond that it seems that LLMs are reaching a point of diminishing returns and the big next move is to make it easy and faster to distill/fine-tune the LLMs for specific business needs, which is something that should be possible to do with the existing infra already (I guess?)
Saying that as someone who keeps my open source projects primarily on codeberg: Getting access to Codeberg CI is a bureaucracy, it has outages due to DDOS attacks every other week and there are a good number of open source developers who are making non-negligible money via GH sponsors.
Sure, for my private projects I already run my own Gitea and Woodpecker CI (and my own docker registry, and my own Taiga server for project management, and my own baserow server to replace airtable, etc...) but the moment you say "just get a VPS to run this service that is available for free at $BIGCORP", you lost 90% of the potential users.
Is it really free, though? You get free service - MS gets everyone's code for free. Only a fool would believe that they don't use private repos for training.
And even if it was free, do you really believe it is sustainable to just offer unlimited service for free to anyone? They've created an environment where you're punished for using anything but github. This is not good.
You don't need to convince me, you need to convince the millions of people who prefer the convenience of "Free as in Beer SaaS" over the resilience and self-sufficiency that we get by hosting our own systems.
I was okay with not having support because I am not part of their customer base. I was okay with not having the webUI, though I wish they made an option where the webUI would be available for some basic-tier paid customers. But I can not be okay with this move. They are just giving the finger to all the community. They never tried to work out a solution that could let smaller users to contribute or support.
I will seriously have to consider moving to Hetzner object storage.