425 karma · joined December 10, 2016
For example, 4chan has some of the most up-to-date English language discussions around for a lot of Japanese fandoms. If you want to discuss new releases of obscure japanese noise bands, you probably want to be /mu/. If you're not on /mu/ and are on some other platform, you're probably just going to get a regurgitated version of the discussion from /mu/ anyway, with an increased propogation delay and someone else doing the cherrypicking. (Note: I'm just using "Japanese noise bands" as an example, I don't know if that particular discussion on /mu/ is good, or if the relevant low-latency discussion is on Twitter, Tumblr, Reddit, etc these days)
If you just link random people directly, you'll inevitably get some folks that can't read the atmosphere and just generally walk all over the flowerbeds without realizing what they're doing. AKA they need to "lurk moar"
https://gist.github.com/allanlw/64aaa1ed1b24fa3eb833da7940b3...
Dockerfile style/feedback appreciated, although this is really quite terse
An interesting property that this paper doesn't consider (I believe the language described does not allow it though) is casting pointers/references to integers. Many languages consider this unsafe, but Rust does not. Considering casting pointers to integers unsafe prevents 'address disclosure' vulnerabilities, which is relevant in the context of any program that incorporates (potentially buggy) unsafe code (read: all of them). This can be used to defeat ASLR.
The new version worked fine for people on 57, but for anyone that was on ESR or a slightly out-of-date version of firefox and updated they got a broken WebExtension that I couldn't push another update to to fix it - Firefox doesn't let you "downgrade" from a WebExtension to a native extension. Thankfully a user emailed me about it and I had it fixed in less than 48 hours, but it was still a mess. All I could do was add a note to the addon description and disable the bad versions and anyone that got broke had to manually downgrade.
The new Webextensions are nice, but it was definitely easy to shoot yourself in the foot with the transition. I know that NoScript also needs the same API that I was using (the ability to inspect response bodies), FWIW.
The extension is https://addons.mozilla.org/en-US/firefox/addon/gpsdetect/ for those curious
EDIT: brevity
This is a very good piece of writing with extensive references and I'll definitely find this useful to share in the future. She documents a ton of tradeoffs she made and resources she chose not to constrain (including the aforementioned syscalls and capabilities), which is important in this type of design and something that I wish I saw more of.
[1]: https://github.com/jessfraz/contained.af/blob/master/seccomp...
For years Facebook (reportedly) did this with hPHPc. Not a very good idea these days, but it's certainly within the realm of possibilities for large companies.
> Poorly tested obscure format support is a goldmine for hackers looking for exploitable code.
IMO the issue is when these code paths are automatically invoked. This is how you get gstreamer running 6502 opcodes for a codec no one has ever heard of when you download a file from the internet.
- HTTP(s) based trackers (although UDP is more common these days) - HTTP webseeds/mirrors (BEP-17) - (if you count it) webtorrent uses Websocket trackers and can support HTTP webseeds (although really of course the P2P is WebRTC)
A small example is that the test cases only work right if English, German, Finnish and French locales are available and if the system timezone is set to "America/Los_Angeles".
For the DS, note that he seems to be targeting the original DS (e.g. not DSi OR 3DS). I remember there being some slightly better stuff but not significantly better than what he made.
I dockerized a working version of HPHPc (which is quite difficult to set up due to bitrot) at https://github.com/allanlw/hphpc-docker if you want to play around with the compiler. I might try to take a shot at passing it through emscripten later.
[1]: http://www.zero-books.net/books/kill-all-normies [2]: https://thebaffler.com/salvos/new-man-4chan-nagle
E.g. if it was linear in that, then you would be going from O(mn) to O(m+n)
It seems that as of a few months ago, two popular (unofficial) command line clients for ACD (Amazon Cloud drive) were acd-cli[1] and rclone[2], both of which are open source. Importantly the ACD API is OAuth based, and these two programs took different approaches to managing their OAuth app credentials. acd-cli's author provided an app on GCE that managed the app credentials and performed the auth. rclone on the other hand embedded the credentials into their source, and did the oauth dance through a local server.
On April 15th someone reported an issue on acd-cli titled "Not my file"[3] in a user alleged that they had received someone else's file from using the tool. The author refered them to amazon support. The issue was updated again on May 13th with another user that had the same problem - this time with better documentation. The user reached out to security@amazon.com to report the issue.
Amazon's security team determined that their system was not at fault, but pointed out a race condition in the source for the acd-cli auth server (sharing the auth state in a global variable between requests...) and disabled the acd-cli app access to protect customers.[4]
In response to this banning, one user suggested that a workaround to get acd-cli working again would be to use the developer option for local oauth dance, and use rclone's credentials (from the public rclone source).[5] This got rclone's credentials banned as well,[6] presumably when the amazon team noticed that they were publicly available.
To top this all off, the ACD team also closed down API registration for new apps around this time (which seems to have already been a strenuous process). I suppose the moral of the story is that OAuth is hard.
[1]: https://github.com/yadayada/acd_cli [2]: https://github.com/ncw/rclone [3]: https://github.com/yadayada/acd_cli/issues/549 [4]: https://github.com/yadayada/acd_cli/pull/562#issuecomment-30... [5]: https://github.com/yadayada/acd_cli/pull/562#issuecomment-30... [6]: https://forum.rclone.org/t/rclone-has-been-banned-from-amazo...