A common example I give is a basic chat client/server, to learn how networking code feels, especially nonblocking IO. Especially if you're implementing the same protocol in the different languages since it means you won't necessarily be as biased by what protocol patterns are easy to implement in one language.
That’s at least 20 ms of round trip time, realistically much more. Especially since it’s not geostationary so the distance to the closest ground station will in most cases be much longer than straight down. I would be surprised if the average ping is below 100 ms.
Latency isn't really the issue here (lots of inference serving happening across oceans, which is probably 100-150ms RT), but bandwidth ends up becoming an issue for large scale transfers (mostly model weights, not the actual inference tokens/queries).
Low Earth orbit is 90 to 120 minutes, depending on altitude. Even if the satellite is currently close to you, it will soon be on the other side of the world.
Assuming that orbital data centers make sense, it will be an interesting challenge to figure out how to best mitigate that. Short tasks that require little data will be easy, but everything else needs rethinking. Can you tolerate variable latency? Should you duplicate the data across many satellites? Or should the data constantly migrate from satellite to satellite?
That's not the reason Cardano failed though. All the capital had already established itself on other platforms and "it uses a better language" wasn't enough of an incentive to switch. It's an overwhelmingly common story in that ecosystem, even with other networks that Solidity can target.
Rust's Allocator APIs have been around since 2016, less than a year after 1.0, they're just now stabilized.
The stdlib types that allocate have been generic (with default generic parameters) over the underlying Allocator used for a very long time. All widely-used data structure libraries where it's feasible/reasonable to use other allocators also are already generic over the Allocator.
It won't "take time to retrofit" because the language and ecosystem have already been prepared for it for 10 years. For context, async syntax was introduced a few years after that and if you just started learning the language today you wouldn't notice it wasn't there in 1.0.
Seems like I've misread the recent Rust version announcement:
> This release brings about the stabilization of the core parts of allocator_api, enabling library authors to define custom allocators that can be used together with certain standard library types - currently Vec and Box. These are enabled through the newly-stabilized Allocator trait.
That made it sound like it was rather limited currently.
I asked an LLM (because it's outside my expertise) and it said:
> It is fair to say that the heaviest architectural lifting is already done: adding a defaulted type parameter (A: Allocator = Global) preserves backward compatibility for existing code, and the core data structures already store and thread the allocator internally. However, plumbing that generic parameter through String, HashMap's std wrapper, and hundreds of standard library methods still represents substantial retrofitting work rather than a completed state.
Looks like The patch notes lead me a bit astray. But it's also not at an 'already all there' state either.
Although it's true that all of Rust's standard library collections are generic over Allocator, Rust 1.100 only stabilizes useful allocator APIs for Box and Vec
If you want to use your MyCustomAllocator with a HashMap or a BinaryHeap or you've lost your mind and want a LinkedList none of that was stabilized yet.
Personally I am slightly interested in HashMap, VecDeque and String but I think Vec is much more heavily used than any of those and will illustrate just how thin the pickings are in this space for real.
Don't diss linked lists. Sometimes they are the best data structure. In my language's interpreter, I maintain "contextual layers" and the best data structure for this is a linked list (two different layers might be applied to the same previous layers)
Perhaps custom allocators could be interesting for BTreeMap and BTreeSet? I am not familiar with their internal implementation, but as I understand it those should allocate fixed-size blocks, which would benefit from a specialized fixed-size allocator.
Sure, in principle all of the standard library collections and indeed most 3rd party collections which allocate could benefit from providing a good Allocator API.
If there's a particular 3rd party collection you know would benefit, you could ask them, or write a PR to make use of Allocator. The standard library collections of course know people are interested, you can think of this as an MVP.
The borrowchecker is not the source of the bulk of the language's slowness. It takes a fairly small fraction of the total time, I think I remember the number was 5%-15% depending on the project.
The majority of time is spent between the trait solver and LLVM optimization passes. There is work being done to improve the trait solver and to produce more efficient LLIR code from the get-go (so LLVM has less to optimize).
Note that is so, because macros can have side effects, that was an oversight of Rust design, not having a way to have pure and impure procmacros (macro rules are pure)
It's also worse because the procmacros are themselves Rust and get compiled with the same parameters as the overall build, so if you're running a debug build the procmacros are running as debug binaries.
> and to produce more efficient LLIR code from the get-go (so LLVM has less to optimize).
I really, really want to have MIR move elimination to run even on debug builds by default. It's the #1 reason for C without optimizations to run (much) faster than Rust without optimizations, and it's terrible
The real solution though is nuclear powered shipping. There's dozens of nuclear powered ships already and they've been running safely for decades. It's just regulatory roadblocks, but if governments wanted to do it they could.
Military ships are "safe", because they can generally defend themselves. Commercial container ships would be the target of hijacking, to steal the reactor / fuel, or use them as 9/11 ship equivalents.
Correct. And military nuclear vessels are also operationally safe because superpower navies are willing to spend a fortune on training and paying qualified nuclear engineers and technicians. Despite those large budgets they still struggle to maintain minimum crewing levels. Merchant lines can't afford the personnel expense.
It has only been two years since the Baltimore bridge was destroyed by that container ship. Had known faulty maintenance, but they kept operating. Now the rest of us have to pay $5 billion to replace the bridge, plus whatever economic damage was done to surrounding industries.
Any sufficiently advanced technology is indistinguishable from magic.
Embed it in the hull and seal it so it’s impossible to remove or open without attracting a lot of attention, while making it call home with a secret handshake every couple minutes so that military can be reach them within hours of an attempted tampering.
It doesn’t have to be perfect, just hard enough to weaponise.
We could find a way to make the reactor so integral to the ship and the fuel so integral to the reactor it’d be almost impossible to remove either. We can compound that with fuel cycles with materials less useful for making bombs.
I doubt merchant ships with nuclear reactors would not be monitored carefully to the point the nearest military would pay a visit if they deviated an inch from their stated planned route.
That and a lot of safety measures integrated into the reactor that would summon military action if anyone disconnects the reactor from any of the various monitoring networks.
Yes I always have to tell it to knock it off, it will ignore memories about it while apologizing for ignoring the memory. I want to know what the edits it's making are.
Modern diffusion models aren't trained like GANs. GANs failed to scale and generalize. Diffusion models do sometimes use Variational Auto-encoders (VAEs), which are a similar structure but which functions differently.
Yes, I’m mistaken. I didn’t realize the ambiguity in the “Florida Democratic Senate candidate” (I misinterpreted it as “Democratic candidate for Florida Senate”).
Yes an extremely contentious race if you pay attention to politics. A year ago if anyone had said Dems have a chance of a Senate majority during the midterms you'd have been laughed out of the building.
A common example I give is a basic chat client/server, to learn how networking code feels, especially nonblocking IO. Especially if you're implementing the same protocol in the different languages since it means you won't necessarily be as biased by what protocol patterns are easy to implement in one language.
reply