new | show | ask | jobs Github

nickysielicki a day ago

For better or worse, pypi is the executable distribution mechanism of the future.

Other cool tools you can install from pypi:

1. https://pypi.org/project/cmake/

2. https://pypi.org/project/ninja/

3. an entire c/c++/zig toolchain: https://pypi.org/project/ziglang/

4. the nvcc cuda compiler: https://pypi.org/project/nvidia-cuda-nvcc/

▲

zahlman a day ago | parent | next [-]

I... really don't know if I'd go that far. Better not to abuse Fastly's good will in providing the bandwidth. These things have PyPI distributions specifically because they support legitimate Python projects. For example, Cmake and Ninja are part of a stack intended to support building things for the SciPy ecosystem, using scikit-build. The CUDA stuff is obviously relevant to PyTorch, Tensorflow et. al. And (per the README) "The ziglang Python package redistributes the Zig toolchain so that it can be used as a dependency of Python projects."

▲

Imustaskforhelp a day ago | parent | next [-]

Obviously a question of good will arises but for what its worth, I wouldnt consider abuse but rather innovation/lets see where the curiosity leads us too

Some time ago in npm, someone has made packages which can install fonts via npm and use the cdn system provided by it for such

I think its more private than many competitors out there. An google fonts alternative is suggested to be coollabs which uses bunny cdn under the hood but using npm's infrastructure which is usually provided by cloudflare is another great idea as well.

Also you are forgetting something that these are economies of scale.

And they aren't using pypi to distribute the official version of deno or the only way they distribute deno. That would be the case which would incur lots of bandwidth good will you could say, but I think the current use case would likely just have in at best 10s of gigs per day or 100s of gigs per day , this is just a method where python is usually installed and it simplifies the installation of deno a lot and there are some really beneficial concepts which can drive up even including recently yt-dlp

its a good idea for what its worth

For context JSdelivr delivered 20,572 TB data per month for free.

I genuinely consider that deno's python might not even reach even 100 GB data per month and I am exaggerating it a lot like with a strech, Python Cuda modules are usually the largest bandwidth eaters imho

All in all, this is an valid implementation/idea. The abuse of good will complaint doesn't stand that much

	▲	zahlman a day ago \| parent [-]
		I'm not specifically objecting to the Deno distribution, but to the idea of PyPI becoming "the executable distribution mechanism of the future". Since the latter makes it sound like people would use it for things that have nothing to do with Python.

▲

data-ottawa a day ago | parent | prev [-]

Totally off topic, the pypi zig library has been very helpful for a few of my projects. It's nice to write low level components and have a simple install process.

▲

goku12 13 hours ago | parent | prev [-]

Why though? What makes PyPI or compatible registries so great for this? On a similar note, Pixi uses the conda packaging mechanism for managing platform agnostic multi-language software distribution. How do these two compare?