Hacker Newsnew | past | comments | ask | show | jobs | submit | Uptrenda's commentslogin

I feel this is a perfect complement to the current 1. link: https://satproto.org/ which implements its own CA system with different trade-offs.

The client fetches the pub key off the server which is decentralized? There's no part in the protocol that authenticates whether or not a pub key is legit. If its replaced by an attacker and someone subsequently goes to fetch a key they can read those messages. I mean, pub key infrastructure is meant to solve that. With SSL and such... that's why you its a federated chain of certificates with providers vouching that names = pub keys.

This is a very common problem. There is potential to possibly make this more decentralized with smart card technology. Like imagine a smart phone with access to pub keys in the hardware tied to an account cryptographically. Then you can say something like phone number = subscriber = pub key. Encrypted messaging apps seem to bootstrap off of ownership for numbers in the mobile system (mobile system security is very bad so there are dragons here.) The other apps like pidgin with OTR plugins they have unique phrases that help with the issue.

When you start looking at decentralized pub key infrastructure tied to human-meaningful names you start to run into zookos triangle:

https://en.wikipedia.org/wiki/Zooko%27s_triangle

human-meaningful, decentralized, secure -- pick two


Yeah, it is a really bad connotation. This infantile and obsessive shallow comparisons of everything. I utterly hate to see the term here on HN.

is this project like vibe coded slop from a zoomer or nah?

Another suggestion to add for you all (IDK how helpful.) When I see PyPy I see that its speed is faster for CPU-bound work but I'm thinking there is also I/O bound work that would see significant increases in the load they can handle. You could host a page that benchmarks common tasks like HTTP req/s (different types) with asyncio vs CPython. Could even have an automated tool that allows projects to benchmark performance from a web-page using PyPi without having to install or measure anything.

Benchmarks are tricky. Do you have a specific use case you want sped up?

I have to say the speed comparison on the front page seems hard to read / backwards

I feel like you should either put absolute numbers side by side or how much faster pypy is (instead of how much time it takes)


It would also be nice to see benchmarks of how much faster PyPy is getting each version. I know there is a tracking page but it tracks dozens of tests and has no absolute reference summary by version.

An easy chart to show v3.x is 10% faster than the last version would be great.


I think the idea is that if an API (or anything really) is using a flexible cryptographic model then it would be possible to string together any system for use in any cryptographic scheme.

For example: you could write the cryptographic equivalent of "if user deposits $100 into this bank account then allow them to redeem these crypto-currencies"

This would mean making systems broadly compatible. Anyone could envision new use-cases for pre-existing systems and be able to extend them without having to modify the existing system. But the challenge of this is (1) designing a simple, fast cryptographic stack (2) widespread adoption.


> For example: you could write the cryptographic equivalent of "if user deposits $100 into this bank account then allow them to redeem these crypto-currencies"

You can do this today.


Thanks for letting me know. I also worked on this in 2013.

Most of the crypto in the OP requires trusted setup phases and is too slow to use for any kind of general-purpose computation. It's the reason why most cryptographic protocols consist of simpler schemes and don't try to do everything. This article is click bait though. Feel like OP just stumbled upon what people have been doing for the past 5 years and wrote this half-baked article on it.

This was true 3 years ago but not generally the case anymore. There's been significant advancements to move away from trusted setups and the speedups with current methods are quickly approaching viability.

List some

0xparc is a research lab. They may be wrong or mislead, but they work on this stuff every day—it's not some guy stumbling across something.

If they are research labs then this article feels like was written by junior researcher they just hired and he is so pumped.

But this stuff is not going to replace current state of things as article claims.


Where are their novel papers?

is there a single thing left that altman promised that he hasn't broken with this company...


Very nice data science. All those studies crunched into a single image. Line length = study number. Green = improvement. It really shows what compounds have the most evidence and effects at a glance.


Thanks :D


^ I think it must have a refer check thing. I went back to the main site and clicked the link from there then it showed the article.


They somehow broke Font Contrast FF addon. If you go to the main site and back it shows the article, but you can't read it without straining your eyes because it is light gray on white. Font Contrast helps, but not in this case. Depending on how I approach to use it, the site either refuses to show the article, or fixes contrast for the heading but not for the bulk of the text.

Some1 worked really hard to break everything.


Not a referral issue. Doing ctrl-click or middle-click or open-in-a-new-tab all are broken too.

Most likely it is just a typical broken spa.

Most web apps, are shut for the websites. They ignore and badly processes the url because things like linking to content, ctrl-click, bookmarks, sharing with friends are afterthought and not on forefront of developers mind.


Yes it is possible to reach the content but navigation is seriously broken.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: