It's a nice idea but the C portion expects an array of unknown size, no bounds checking possible at all and it's also not returned how large the PNM image is. I also don't understand why glib is needed here to get the array to work. It's also missing the stdint.h include to be able to use uint8_t.
You're missing the point. Volatile forces two loads of a value that may have changed in the middle. So the value of "x" may depend on the time/order of load.
Which is, if I understand correctly, the entire point of volatile. Don't use it if you don't want that behavior.
And in fact, in the example given, if there is something (another thread or whatever) that can change the value of x, then you don't know what either number will be. Well, in that circumstance, without volatile, it may print the same number both times, but you still don't know what the number will be (unless the read gets optimized away entirely).
I suspect that many undefined behaviors reflect the inability of the standard committee to come to a consensus on the nuances involved. “Punt to the implementers” is a way to allow every tool vendor to select their own expected behavior in those cases.
You seem to be operating under the assumption "undefined behavior" means "the compiler authors can decide what to do." That's not what it means. It means "any program that causes this behavior to be triggered is not a valid C program, the programmer knows this and did not submit an invalid program, and the programmer explicitly prevented this from happening elsewhere in ways automated analysis cannot detect. Proceed with compilation knowing this branch is impossible."
The spelling for compiler authors getting to choose a behavior is "implementation defined", as the other comment mentions.
It means the C standard does not specify what the program does. Other documents may still specify what the program does. And the program definitely still does something, whether specified or not.
Why is that missing the point? Loading it twice, possibly with different values, is the intended behavior. It's only undefined because the C spec doesn't specify the order of the loads (unlike most other languages which have a perfectly well-defined order for side effects in a single expression).
No I'm just repeating what the original comment said, which is that it's explicitly UB:
"5.1.2.4.1 says any volatile access - including just reading it - is a side effect. 6.5.1.2 says that unsequenced side effects on the same scalar object (in this case, x) are UB. 6.5.3.3.8 tells us that the evaluations of function arguments are indeterminately sequenced w.r.t. each other."
If function arguments were sequenced with respect to each other, it wouldn't be a problem.
But actually, maybe the original comment is wrong. Presumably "indeterminately sequenced" and "unsequenced" mean different things, although I don't have a copy of the standard at hand to check.
Yeah dude but you've really marketed D poorly. I remember looking at D what must be 15 years back or so? And I loved the language and was blown away by its beauty and cool features. But having no FOSS compiler and the looming threat of someone claiming a patent (back then it was unclear that Mono/C# was "legal" and even Java hung in the balance) was too scary for me to touch it.
Now I'm old and I believe D has missed its opportunity. Kinda sad.
Open source and free software isn't the same thing. Nobody made a claim on Java either, until someone did. I just distinctly remember explicitly not exploring D for that reason. Also this way way before LLVM and I also don't think GNU had a D compiler back then. There was only the (and I really believe it was closed source) Digital Mars compiler.
15 years ago, both LLVM and GNU had a D compiler. gdc (the GNU compiler) was not an official part of the gcc collection, but it was definitely there and 100% open source.
All three compilers shared the open source D front end. The DMD backend source code was available for anyone to use, it just couldn't be redistributed. We were eventually able to fully Boost license it.
The DMD compiler always had source available for free from Digital Mars. I never sold a single copy :-)
My point is, in your exact example both reinterpret_cast and C-style casts have the exact same behavior, making the example bad. If you want to showcase a deficiency of C++, it would make sense to pick something where the difference between cast types actually matters.
...and an incredibly handwavy shallow explanation of why this actually works ("Through a clever sequence of oblivious transfers and what’s called multiplicative-to-additive share conversion, they each compute a partial signature.")
I don't get it. If you want a blog, write a blog. If you don't want a blog, don't write a blog. But why use an LLM to create a slopblog? It just wastes EVERYONE's time and energy. How disappointing.
Not sure if it's AI slop yet, but I also found the core part (the "oblivious transfers") to be explained too handwavy to really understand the properties of this system. I don't want to know all the mathematical details, but I do want to understand who is exchanging what data with whom. "oblivious transfer" doesn't tell me anything here.
The other (maybe more interesting) question is how this tech would be deployed. So ok, we have a system, where something can only be signed/decrypted/encrypted/etc if several parties are in agreement. Who should the parties be? How is the threshold itself actually managed?
OP also seems to drift between different usage scenarios here:
- some sort of collectively owned good (like the DAO or resources in a cooperative?) - seems straightforward on a technical level (every owner has a partial key) but also a niche usecase and quite inflexible: What happens if an owner drops out or you want to introduce a new one? What happens if you want to change the quorum?
- traditional authentication of individual users against a server, in a federated setup like the fediverse: Seems like the most practical usecase: One party is the user, the other is the server, the verifying party would be other servers of the network. But then you have to pick your poison by how you set the quorum: Either the quorum is "any party can decrypt the data", at which point you're not better than normal password auth; or "both parties are needed", which would protect against the user or the server accidentally leaking the key - but then you're back to "single point of failure" if any party accidentally loses the key.
- the last scenario would be server-side keys that could cause massive problems if they leaked. But I don't understand at all who should be the other parties here. Also how would this be better than HSMs?
Oblivious transfer - party A creates two random values (x_0 and x_1) and sends them _both_ to party B.
Party B picks one and uses that to compute future values that are sent back to party A _but without telling party A which of the two values they picked_.
In this example I'm hand-wavey because the production math is complicated and confusing - I took a vastly simplified approach that still works functionally for the demonstration without fully implementing the OT protocol.
> what happens if an owner drops out or you want to introduce a new one? what happens if you want to change the quorum?
In either of those scenarios, assuming you still have quorum, you can regenerate keyshares for the new group for the same public key (and underlying yet unknown private key) by re-running the ceremony with the new participants. Production implementations of the protocol fully flesh this out.
> traditional authentication ...
I wouldn't use TSS in that setup. Traditional auth + MFA is more than adequate. The better use case would be where you have a group that needs to demonstrate consensus (like governance for a programming language, multiple parties involved in signing an application release, or even an HOA that needs to vote on policies). In all of these, you'd take an M of N approach (rather than the simplified 2 of 2) for achieving quorum.
Yeah, AI blogs are close to worthless. It’s a circular feed of slop for LLM’s to be trained on. If I can just talk to the LLM to get the same content, I don’t want to be directly reading it.
What I want to read is well-researched and deeply considered pieces that do a good job explaining concepts in a fresh way and help me learn something new. Sure, use AI to help get there, but if you haven’t done much research or haven’t thought about it yourself beyond the prompts… I don’t want to read it
