Hacker Newsnew | past | comments | ask | show | jobs | submit | more BiteCode_dev's commentslogin

Gravity is not instantaneous.


The article is well written, and a good history of the all ordeal. But please note it gives more weight to the "against the GIL" side of the story, and all that can go wrong.

It doesn't highlight enough the other side of the coin:

- Sam's work is very high quality, and he brought with the no-gil some unrelated perf improvements so that people don't feel like they loose perf too much.

- Sam played the open source game perfectly, and is incredibly patient given what he is bringing on the able and how slow and flaccid the steering council reaction was (without the community pushing on it, it would still be collecting dust).

- Sub interpreters have yet to demonstrate any usefulness at all in Python. In fact, any serious metrics at all. This is the first attempt to be that well defined, and measured.

- The community feedback shown a great interest in this particular project.

- The steering council did conclude "We intend to accept PEP 703, although we’re still working on the acceptance details."

I'm not a no-gil enthusiast. I would be fine with have it never be removed, and I think we should try sub-interpreter first.

But what's fair is fair.


> I would be fine with have it never be removed, and I think we should try sub-interpreter first.

There is a lot of work going on with sub-interpreters and the per-interpreter GIL is shipped in Python 3.12.

The results are very impressive: https://lwn.net/SubscriberLink/941090/8bcb029dbf548f26/, as good as one could have hoped I think.

It seems to me like the work on sub-interpreters will continue in parallel ;) to the work on free-threading.

Sub-interpreters and no-GIL have different use-cases though.


For now there is nothing impressive with it. The perfs they show off don't include data sharing, which is very primitive.

Which means there is nothing subinterpreters do that can't be done with multiprocessing.

They show progress, and it's good. But I would wait until we can see gunicorn spawning WSGI interpreters workers and get similar performances to the setup with regular workers to get enthusiastic.


to further your point: I've used the technique launching a separate python process for performance critical operations for years on different projects. I can run it in a completely different priority set as well.


Once no-GIL is integrated, what is the use case for sub-interpreters?


I have a use case for sandboxing (without a secure-sandboxing requirement) that it would help solve. But at last read, they discourage/disallow sharing modules across the subinterpreters and reloading all modules in each subinterpreter is not acceptable for my use case.


That's very interesting, thank you for the link. I've used Python since the 00s and I don't ever recall having come across sub-interpreters...


They were always usable using the C-API but not exposed to the Python library (and until 3.12 you were blocked by the shared GIL so there was little point anyway).

Now that the work from Eric Snow has been merged, you can use https://pypi.org/project/interpreters-3-12/ to create one from Python code.


I'd be interested in introducing appropriate abstractions (ABI) and in a module for parallelism. Whereby one can swap out different implementations for sub-interpreters, green threads, GIL or even hardware or cloud vendor-optimised implementations. Not that dissimilar to Project Loom in Java.


PEP582 - “Python local packages directory” is my personal bugbear when it comes to Python annoyance: https://peps.python.org/pep-0582/

I believe PDM still supports it. I just hate virtualenvs for builds and deployments and wish Python could just do what JS does. It’s been proven that it can. https://discuss.python.org/t/pep-582-python-local-packages-d... Is the discussion thread. Also frustrating that “exactly one correct way to do something” seems to be one of the justifications thrown around for rejecting this when I’ve never found that to be true in Python.


Yeah, python is great but I wish package management was more standardized and simpler.


What I’d liked to see is the performance patches without the no-Gil part.


The performance patches have already been merged and there is ongoing working in the Faster CPython project.


You're already using a bunch of those in recent Python versions.


SO, why would people be against removal of GIL ?


Here are some possible reasons.

They use unmaintained C extensions that won't be updated.

They maintain C extensions that would be complex or painful to update.

They're concerned about fragmenting the ecosystem into GIL and No GIL.

They think that it will make single-threaded Python programs slower.

They have no interest in multithreaded Python and don't think the additional complexity is justified.


> They use unmaintained C extensions that won't be updated.

Well, that's on them.

> They're concerned about fragmenting the ecosystem into GIL and No GIL.

Same way we have fragmentated ecosystem into ASYNC vs SYNC now? /s

> They think that it will make single-threaded Python programs slower.

Python is already horribly slow. A little bit slower single thread programs shouldn't be a problem. But the inability to run multi-threaded programs effectively in this day and age is very much a problem.


I haven't fully confirmed this but my understanding is the proposal forks the ABI into two versions. which IIUC means any package with native extensions needs to release 2x as many wheel files (not to mention if I need to have two versions of the interpreter installed to build them, not sure though). seems like a mess


We had an client side pastebin called 0bin.net

Recently, the last CP addict idiot basically killed us. We always remove those posts when we get a report (we can't see them by our-self, that's the point), but we were swamped by other things, and we didn't catch this one in time. It's a free volunteer project after all.

Our host took down our server and our backup server. We didn't have an offline backup because it makes taking down bad content more work.

We are still struggling to go back up.

This is why you can't have nice things.

(and also, don't host your backup server with the same host)


I also appreciated 0bin.

Internet progress seems to resoundingly be going backwards. Remember when pastebin used to be super active? Hell, remember when google had informative search results?


Internet, computing, etc.. As time goes on and attestation catches on general purpose computing will die for the most part.

But maybe we'll see a renaissance of sorts? When the only sites that work on Firefox are fediverse and a few hobbyist sites and people need to go out of their way to find computers that they actually control we might undo the eternal September of the web as a whole.


All is part of the capitalization of the internet. People seek to make money and how they do it is not always morally or ethically sound.

The view of “the global village” with betterment for human society has become another rent seeker “getting my own” hustle culture. It’s the way people are I guess on the whole.


It's not as simple as capitalization, we had a long period of time where people were making fortunes but the tools actually worked.

Theres something deeper thats causing things like Google search refusing to show more than 12 or so results and them being lower quality. Causing YouTubes algorithm to recommend things you've already watched or just recommending a video over and over for weeks until you watch it to get rid of it.


Yeah!

I guess the thing is that the push is on for ever higher extraction and yields for managers/stakeholders/CEOs and to show it to shareholders and investors and the like. And with higher interest rates where money costs its only amplified I reckon.

I promote a video ad campaign for a client and its insane how much it is being shown for a few clicks. Ad revenue is probably going down in relative terms (or more effort for same results)


> Theres something deeper thats causing things like Google search refusing to show more than 12 or so results and them being lower quality.

That something is - wait for it - capitalization.

Google, YouTube and the rest of the Internet are drowning in low-quality, automated, yet monetized dreck. The signal-to-noise ration on the web has never been lower, and it's getting lower still.


> progress seems to be going backwards.

I blame JavaScript.


And chewdata, member chewdata?


For what it's worth, 0bin is an excellent service.

I hope you figure things out in a way that causes less of a headache for you.


> and also, don't host your backup server with the same host

cheap way to get some resiliancy is to host a hot-spare/replika at another provider/isp and failover at the dns level with an alternative record


I very much enjoy using 0bin, its excellent. Hope you figure things out!


0bin is a wonderful service and I really hope you're able to get back online soon. Never change!


This is a big barrier to starting cool websites. Perhaps a local offline AI instance with its own keys could help?


Sounds like what really killed the service was your inability to deal with reports in a timely manner.

If you can't promptly handle take down requests don't put yourself in a position where you're required to promptly handle take down requests.


Right, so like hosting your server in a jurisdiction where one can ignore take down requests, such as places like Albania?


I'm guessing every language has a module to abstract that.

E.G: for python, the current go-to pypi module for this is platformdirs (https://pypi.org/project/platformdirs)

It's probably easier to promote the use of the module than explain the standard.


Would love to see examples for Rust, Go, NodeJS, C# and Java so that I can add them to my GitHub stars. I could find a few results for XDG support in Go, Rust, and NodeJS, but none of the aforementioned languages supported standards across all of macOS/Windows/Linux. Hopefully it's just my Google-fu failing me and people can correct me in the comments?

C and C++ still have strong roll-your-own culture, with the exception that C++ now has a decent filesystem module in the standard library but still no standard config file parser AFAIK, everyone usually imports nlohmann's JSON library.


For Rust, use the "directories" crate: https://crates.io/crates/directories

That will respect platform conventions on Windows, macOS, and Linux/UNIX.



Many of the programs complained about are written in C. I'm sure that there are libraries for c that abstract this, but there isn't one that is common enough that most c programs can depend on it.


And hence the numpy API.


Indeed, but this general wisdom is too vague to be useful on its own.

When I was young, I heard it many times, and couldn't make anything from it.

What's the audience exactly? How do I know them? Once I know them, what am I suppose to do? And all that stuff is contextual, so what are we talking about really?

When learning to negociate, knowing your audience is a very different thing than when trying to teach. Very few people are able to pull it off accross fields.

So I decided to write an article that targets a specific case with a concrete application and examples.

This would have been more useful to me back then.


Lot of comments here are replying the nature of the solution matters.

E.G:

"Sure you are being paid for a solution. But you are also probably being paid for a quality solution"

or

"Code can be the solution. Good code even more so"

or

"when you pay to get your car fixed, there's an implicit expectation that it will be fixed well enough that it doesnt break the same way again"

But the article does address this:

> Now, of course, this is a tad more subtle.

> Different types of problems have different types of costs, consequences, and constraints. And solutions to those problems as well also have them.

> When code is the solution, it has a cost, at the very minimum in time and man-hours. And constraints, such as licenses, NDA, platforms, versions, formats, resource limits... And of course, consequences, such as technical debt, electricity consumption, user experience and hiring difficulties.

> While the clients don't care about the code, they care very much about those.

> And so, as IT professionals, we must communicate talking about that. Not about the code.

Of course the quality of the solution matters, but it any good thing has a price. Therefore you have to present it with the price, and then your opinion on the whether it's worth it.


I was pleasantly surprised to realize my little part of town is getting quieter and quieter because more and more EV are getting in.

It's a delight.

Even if it were polluting just the same forever, it would be a significant improvement.


And usually they are combined with existing parking spots.


AKA encryption.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: