Youtube comments of TheEvilCottonBall (@alexandermaximilianoetken7265).
-
58
-
31
-
> i think developers do technically have a choice, they could go use directx and metal instead, both of which are much more cringe masochism than vulkan
Direct X and Metal are tied to the Windows and Apple platform and Vulkan is the graphics API of Android and Linux. You do not have a real choice, you have a choice of what graphics vendors provide you. I cannot write my own OpenGL implementation or an entirely different API because GPUs are closed, undocumneted hardware, so there is no big competition in terms of graphics APIs. Yes you can layer stuff on top of other APIS (with the associated drawbacks) like Vulkan on MacOS is, I think, but there is no level playing field for graphics API implementations.
> a lot more refreshing to use than something like springboot or react
These are not graphics APIs
But I agree, an even better example for libraries that have no competition would be system libraries/APis (like win32 API, Android SDK, linux syscalls, libc, etc.). They will have to be used (directly or indirectly) no matter how good or bad they are.
26
-
24
-
22
-
20
-
19
-
17
-
16
-
15
-
11
-
10
-
9
-
8
-
7
-
7
-
7
-
7
-
6
-
6
-
6
-
5
-
5
-
5
-
4
-
4
-
4
-
3
-
3
-
3
-
3
-
3
-
3
-
Well, if your Hello, World program runs on top of a buggy OS then it is also crap, you need to run on bare metal to have ideal software. Also your hello world program likely relies on a compiler, assembler or interpreter (maybe a hex editor). Therefore it is not immune to compiler/...-bugs. I am curious, what kind of error does your program output if it cannot write to stdout e. g. if its output s piped to /dev/full does it write to stderr and exit with a nonzero exit code. This is a common bug in Hello World Applications. Does your program rely on libc stdio? Well, then you may have bugs/different behaviour on differebt systems. Then there is the question what ideal software really is, a Hello World App which does not offer different languages, Unicode, RTL maybe not ideal for international users. Ideal depends on the user, thd use case, and so much more that it is very hard to ever claim that a piece of software is ideal in a geberal sense.
3
-
3
-
3
-
3
-
3
-
2
-
2
-
2
-
2
-
2
-
2
-
2
-
2
-
2
-
2
-
2
-
2
-
2
-
2
-
2
-
2
-
2
-
2
-
2
-
2
-
2
-
2
-
2
-
2
-
2
-
2
-
2
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
I wonder how Mojo compares to Cython. Cython is also a superset of the Python language intended to run pythonic code at C speeds. However Cython markets itself as an option for scientific computing whereas Mojo markets itself as a language otpimised for AI and ML computing. Of course Cython has been around for longer and is more mature. Mojo for instance does not even run on top of Windows yet (only WSL) and isn't ready. I really like some ideas about Mojo autotuning and Rust like ownership semantics, not found in Cython. I guess we will have to see, how compatible/performant Mojo really is and how it compares to other fast Python solutions: lpython, pyston, mypyc, jax, numba, cupy, taichi, PyPy, triton, nuitka, brython, skulpt, pythran, weld, ironpython, transcrypt, pyccel, pyjs, grumpy, uarray, shedskin, jython, compyle, codon, seq, HOPE, transonic and many more. Some of them are incompatible with CPython and therefore limited in their use e. g. Codon. Some limit themselves to only a few jitted, performance critical functions e. g. numba, some try to be fully compatible with Cpython e.g. PyPy, nuitka which only brings modest performance improvements so numerical stuff in PyPy-Python is slower than calling C libraries within CPython, and some extend the python language e.g. Cython, Mojo. The problem with the superset approach is that you are kind of writing C in Python, so why not write in C instead?
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
1
-
@nagualdesign It wont land on the moon, if SpaceX cannot get the lander developed. Starship is an increadibly hard design, (full reuse requires high engine performance, getting to the moon requires cryogenic propellant transfer between multiple ship, SpaceX needs to relight a Raptor engine in Space after the propellant depot has been in orbit for several weeks if not months, they need to think of backups in case their stupid cable crane fails, they need to develop lunar Starship itself, land it on the moon to prove that it is worthy and so on... They can fail in many ways, and Starship is a bit like the Cybertruck, costly to develop, eventually scaled back to be more traditional in manufacturing, no real market (few people need 100 tons to LEO in rapid cadence), economically suboptimal (Starship can currently maybe get 40 tons to orbit (they still fly empty without mass simulators, so maybe even less) and still has problems with re-entry. Well, we'll see how it goes, maybe the chinese will be first this time...
1
-
1
-
1