It's something I've seen countless times in my life. A piece of software achieves something pretty close to perfection for its task, but the self-imposed "need" to continue development on it ruins it.
But I keep getting told I need a "Slackware intervention" :P
Totally unnecessary.
I now try my best to avoid anything that involves Python.
With your argument about compatibility no programming language would ever be able to evolve or deprecate/remove some features. Look at how many things changed e.g. between recent C++, Rust or C# major releases. Python had only one major revision in 15 years in comparison!
If you are using software that still requires Python 2 that's very much a problem of that application and not Python. Complain to the vendor responsible, not about Python. They had ample time to update their code.
End users had no business touching Python 2 for at least a decade now. And upward compatibility between the point releases of Python is (and has been) generally pretty good.
> With your argument about compatibility no programming language would ever be able to evolve
"evolve" and "change" are not ends in themselves. If it was usable as it was, then there was no need to evolve or change in an incompatible way.
15 years ago: also irrelevant. World War One was unnecessary, and that was 109 years ago. Whatever issues existed are sorted out by now, but that doesn't mean it had to happen.
That you find them unnecessary because the original code was "usable as is" for you is irrelevant.
The reasons why these changes have been done (for a developer they have been fairly minor) has been hashed and rehashed for the past 15 years, Guido van Rossum wrote on it extensively too.
Most has been performance-related (Python 3 is significantly faster than 2) and probably the biggest user-visible change is the clean up of string handling with everything being Unicode now.
Frankly, anyone complaining about this stuff today is just beating the old dead horse for the sake of having something to complain about. The same like some people constantly bringing up the Linux systemd discussions after more than a decade, even though they are completely irrelevant today.
1. It's not true
2. It's true, but it's not important
3. It's true, and it's important, but we're tired of talking about it.
I'm not lying. This problem cost me a full weekend earlier this year. I may have misremembered the exact version numbers, though. Perhaps it wasn't between 2 and 3. Is there a Python 4?
The problem is that you can't have one Python interpreter that covers all versions of Python, and Python is very unforgiving about using the wrong interpreter version.
In any case, it was a serious problem that cost me a lot of time and really soured me on Python as an end user. I've not encountered a similar issue with any other language before, so this appears to be uniquely a Python thing.
But I don't know. I'm happy enough to just avoid Python-based anything whenever possible now.
If you have downloaded Python 2, you must have done it intentionally, it isn't even available from the Python.org website download anymore - and had not been for a long time.
>The problem is that you can't have one Python interpreter that covers all versions of Python, and Python is very unforgiving about using the wrong interpreter version.
Because it doesn't make sense. The language evolves. Nobody is going to stop working on it only because it could break some old code somewhere. The changes from 2 to 3 happened *15 years ago*.
There is no reason to try to compile/run old Python 2 code today - and if you still do need it for some reason, then you need to download the Python 2 version (which are still available, if you need them but one has to look for them). But then you better know what you are doing.
If some old code requires Python 2, any somewhat experienced Python developer will spot that right away (e.g. the use of print statement in Python 2 vs. print() function in Python 3 is a dead giveaway).
>In any case, it was a serious problem that cost me a lot of time and really soured me on Python as an end user. I've not encountered a similar issue with any other language before, so this appears to be uniquely a Python thing.
I don't doubt it has costed you a lot of time but this was a problem entirely of your own doing by not doing your homework.
You would have exactly the same problems if you tried to compile old K&R C code or C++ code from 20 years ago using modern compilers. Or tried to feed modern C# code to a compiler from 10 years ago. Or, God forbid, tried to run some modern Javascript code using old browser - or that old HTML with Flash and what not in a modern browser ...
Python is actually much more lenient in this regard because a major language change has happened only once, 15 years ago, with Python 2 being deprecated for well over a decade. The point releases are all upward compatible with no issues.
> You would have exactly the same problems if you tried to compile old K&R C code or C++ code from 20 years ago using modern compilers. Or tried to feed modern C# code to a compiler from 10 years ago.
If you feed 10 year old C# to a modern compiler it will compile it fine. I’d imagine most C and C++ of even 20-year vintage will still compile. The issue is of forward compatibility, not backwards.
Likewise C and C++ have had enough breaking changes since 2010.
With that said, neither IL nor C# itself had any changes breaking forward compatibility (aside from a very early change to foreach recently discussed here).
Also, "it depends" isn't the same as "If you feed 10 year old C# to a modern compiler it will compile it fine."
If you're shrinking then a competitor is providing better options, or your problem space has shifted.
Various dependencies, like XML parsers, JPEG decoders, HTTP libraries, etc, change over time, drop deprecated APIs and need the main application to adapt. It'll also bump into the tech changing around it -- UTF8, year 2038, 64 bit CPUs, etc.
For a project this size it takes work just to stand still and keep it comfortably buildable on a modern Linux distro.
death is part of life though, and the only thing that grows without and end is cancer (quite literally).
I don't see why project can't just be considered "done": yeah you update dependencies and make it work with new/current formats, but essentially it's done. The use case is now clearly defined, implemented and tested.
AWS offers hosted versions of each. They are not, as you say "rebranded and for sale", but rather are hosted services, running whatever upstream releases.
But to say that the project as been absorbed by AWS (in either case) is simply false. Spark is developed by a community of numerous companies, of which Databricks is usually most prominent. Kafka is likewise a community of several organizations, of which Confluent tends to be the most influential.
Yes, Apache is (and always has been) business-friendly. The license enables companies to profit on those projects. But, for the most part, companies that benefit from those project also contribute back to them, as a way to ensure sustainability. That's how it is supposed to work.
Apache httpd certainly helped tons of companies profit, so helping companies profit doesn't seem off-brand.
Your comparison to Terraform is also interesting. Apache Kafka, for example, has been absorbed by Amazon (MSK). Apache Spark is Amazon EMR. Just looking through Amazon's paid service offering I find several Apache projects rebranded and for sale. A cynical take could be that Apache is an org that helps companies like Amazon profit off of the work of open source developers.