Hacker News

Top stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. Human brain is two separate organs, Stanford Medicine-led research finds(stanford.edu ↗)
    74comments
  2. If math is more than proof, we need to better celebrate the rest of it(terrytao.wordpress.com ↗)
    40comments
  3. GPT-6 Astra Solves a WWI German Radio Cipher(prinzai.com ↗)
    42comments
  4. San Francisco Onion Futures Company(onionfutures.com ↗)
    62comments
  5. Android 17 is the first since 3.x to add new APIs without releasing to the AOSP(grapheneos.social ↗)
    404comments
  6. Apple M6 Pro Achieves the Highest Single-Core CPU Score in Geekbench 7(geekbench.com ↗)
    15comments
  7. Typesafe-computer-use drives a Mac toward a goal for 1/50th of a cent per step(github.com/awlevin ↗)
    43comments
  8. Cloudflare Quick Tunnels(cloudflare.com ↗)
    278comments
  9. SDCC – Small Device C Compiler(sourceforge.net ↗)
    17comments
  10. Science Is Open Software(jepedersen.dk ↗)
    36comments
  11. You can run Git on object storage if you re-make packfiles(tigrisdata.com ↗)
    12comments
  12. Why building a Rust LSP is hard(rust-glancer.github.io ↗)
    31comments
  13. How to Write with an LLM(sockpuppet.org ↗)
    324comments
  14. Saving another 100TB of RAM(cloudflare.com ↗)
    70comments
  15. Ctenophores: Wonders of Biology(quantamagazine.org ↗)
    5comments
  16. How OpenAI Used Its Own LLMs to Design Its Jalapeño Chip(ieee.org ↗)
    85comments
  17. NASA-IBM Lunar Foundation open-Source Geospatial AI Model(usra.edu ↗)
    2comments
  18. The first new cat species discovered in 100 years(nationalgeographic.com ↗)
    104comments
  19. OpenJev(openjev.com ↗)
    262comments
  20. Goroutine Leak Profiles(go.dev ↗)
    2comments
  21. Show HN: Cactus Needle 3: 8-29MB automation models can match DeepSeek V4 Flash(cactuscompute.com ↗)
    88comments
  22. Photon-Emission-Guided Laser Fault Injection Enables RP2350 Secure Debug(ledger.com ↗)
    69comments
  23. Minimal Phone 2(minimalcompany.com ↗)
    220comments
  24. Cache-to-Cache: Direct Semantic Communication Between LLMs (2025)(arxiv.org ↗)
    12comments
  25. Veronese's Dogs(publicdomainreview.org ↗)
    discuss
  26. Suppress vulnerabilities applying Kubernetes context to scans(github.com/alegrey91 ↗)
    1comments
  27. Warez: The Infrastructure and Aesthetics of Piracy (2021)(archive.org ↗)
    61comments
  28. Inside ZCode: Silently uploading your Git history to the cloud(ferstar.org ↗)
    98comments
  29. Cyclomatic Complexity in C#(ndepend.com ↗)
    19comments
  30. Stepfun Step 5 Preview (LLM): On AA Pareto frontier(artificialanalysis.ai ↗)
    2comments

Combating abuse in Matrix – without backdoors

222 pointsby 5y agomatrix.org
18 comments
5y agoHN ↗

I'm surprised the site doesn't convert a dupe post to an upvote (if the user hasn't already upvoted).

5y agoHN ↗

It does, but only in a certain window (I believe 12 hours). After that, dang and other admins can (and have done) merge duplicate threads.

5y agoHN ↗

It does, based on a URL match. One of the links included a "www." hostname prefix in the URL and the other did not, so the URLs don't match.

5y agoHN ↗

So HN could detect dups based on hashes instead? I understand the HTML could be different while the content is the same but this is an extra step that helps.

5y agoHN ↗

Then you'll have one of these "fancy" modern websites that load their content entirely through js and it'll fail because the HTML is the same on all pages.

A URL match is probably good enough the vast majority of the time. Maybe it could also support a bit of fuzzing, such as matching with and without the leading www and both http and https. Beyond that it's probably asking for trouble.

5y agoHN ↗

Then you'll have one of these "fancy" modern websites that load their content entirely through js and it'll fail because the HTML is the same on all pages.

What are you talking about, it would be done entirely on the backend.

Mind you, I've brought it up with dang multiple times and he says it would be a hassle and too brittle to be effective (fair enough), but nothing about it would require javascript.

5y agoHN ↗

I believe the parent is talking about submitting links to websites that render their content via clientside JavaScript and how that would break the hash dupe detection. They aren't suggesting that the functionality would need to be implemented in JS by HN.

Regardless, hashing the content to detect dupes is just an idea that wouldn't work for a lot of reasons.

5y agoHN ↗

one of these "fancy" modern websites that load their content entirely through js [will] fail because the HTML is the same on all pages.

That's a feature, not a bug. (Although it would admittedly be better to block those explicitly rather than relying on coincidental interactions with something that doesn't seem directly related.)

5y agoHN ↗

hashes of what, though? Without accessing the content, the URL is essentially the only thing to go on. Plus having the same post make it to the front page pretty much means the system is working exactly as intended: content that HN users are interested in makes it to the widest audience (not everyone checks HN multiple times a day =)

5y agoHN ↗

THANK YOU. I have never understood why folks on Reddit, etc. are so vehemently opposed to reposts. If people are upvoting it, that means they like it. If they like it, that means either a) it's new to them, or b) they enjoy seeing it again.

5y agoHN ↗

For what it's worth, my comment wasn't meant as a complaint that it was posted again. I just wanted to make sure that folks saw the previous submission, since that one has comments by the Matrix lead.

5y agoHN ↗

Understood, and I (for one) interpreted your comment as you intended it (and - thank you for providing that context!).

5y agoHN ↗

Who is going to crawl all posted links? Also, HTML keeps changing. Unnecessary (and very bad) way to check.

5y agoHN ↗

We tried that kind of thing and it was a nightmare. Trying to make general content-processing things on the web is a full time job and more.

I do think it's practical for us to make use of <link rel='canonical'> though. But many pages don't include that. The OP doesn't for example, so that wouldn't have helped here.

5y agoHN ↗

Matching against the title and domain would probably catch most duplicates, including this one. Although that's still work.

5y agoHN ↗

There are so many small variations in titles that one probably couldn't rely on that alone, but it's a good idea.

5y agoHN ↗

Comments merged thither. Thanks!

other than these of course