6 ms·

Py3k(python3) more than one year on - 0.96 % packages supporting py3k.

16 points by illumen 17y ago

brazzy 17y ago
It's often said that Java has been held back (feature-wise) by Sun's obsession with downward compatibility. This article indicates that it may be preferable to the alternative... What's the point of new features when hardly anyone uses them because not only their own code but also most of the important libraries don't support the new version?
- chaosmachine 17y ago
  Drupal suffers from this same problem. Every time a new major version comes out, all of the existing modules break and have to be rewritten. With over 4000 modules (and roughly the same number of module authors), it's a serious problem. What ends up happening is everyone keeps using the old version of Drupal until they're forced to upgrade or abandon it due to security concerns (old versions don't get patches from the security team).
- steiger 17y ago
  It's of course a trade-off. I think the changes that made into Python 3.0 were welcome, and breaking downward compatibility was a hard but acceptable consequence. It's a difficult matter: Let the language evolve and break things, or make it grow old and stale? Careful ponderation should be made.
- amix 17y ago
  I don't think backwards compatibility is the problem. The problem with Py3k is that it does not offer any vital features, it requires a lot work to port to Py3k (to top that off Py3k is slower than Python 2.x). The only way I can see Py3k successful is if there are clear advantages to move to Py3k, like improved performance or extra vital features (like improvement of inline functions (lambda)). The improved performance is also a big question as teams like PyPy and Unladen Swallow are working on improving Python 2.6 and not Python3k.
  - eru 17y ago
    > extra vital features (like improvement of inline functions (lambda)) I would not consider this a vital feature. You can already nest arbitrary functions, you just have to name them. However Tail-Call-Optimization would be a vital feature. (Alas, it's not going to happen.)
    - masklinn 17y ago
      I'm pretty sure the improvement he was talking about is "full" anonymous functions (w/ multiple statements & everything).
      - scott_s 17y ago
        And I'm pretty sure eru understood that, which is why he pointed out you can achieve the same functionality by naming them.
        eru 17y ago
        Yes. And if you want multiple statements, you are either doing site-effects (evil) or using variables inside your closure, so you might as well name it, too.
        masklinn 17y ago
        > you are either doing site-effects (evil) Uh not necessarily, and even if you are there are many situations where it's interesting and useful. > or using variables inside your closure, so you might as well name it, too Does not follow.
  - jnoller 17y ago
    How about this one: 2.x is dead. While bug fixes and minor changes are occuring, it's going to slip into maintenance mode once 2.7 comes out. All the cool things, such as importlib and other items are py3k only.
- masklinn 17y ago
  > This article indicates that it may be preferable to the alternative... No. Python 3.0 was never expected to see widespread adoption, nor was it expected that libraries be ported to it. Ever. Python 3.1 is expected to have a better adoption rate, and library porting to start in earnest during its time and even more with Python 3.2 next year, especially with tools such as 3to2 appearing. Everything is going according to the plan, and this plan is the reason why 3.1 was released less than 7 months after 3.0 (3.2 will probably be released early next year, putting barely a year between 3.0 and 3.2). That's also why 2.7 is in the works (and was already planned before the 2.6/3.0 releases), and there might even be a 2.8 (though that's unlikely). edit: when you get down to it, those claiming Python 3 has failed are either morons or FUD-spreader, Fox-style. I'm not saying Python 3 has succeeded, but we won't know whether or not it has failed before the end of 2010 at the earliest. Right now, everything's going according to the plan. Please remember the planning for Python 3000 started for real during the 2003 State of the Python Union. Meanwhile the release of Python 3.0 was in december 2008.
  - jnoller 17y ago
    I love you
    - masklinn 17y ago
      I love you Well I don't know if I love you but you sound like somebody I could love.
garnet7 17y ago
Python 3 can't just be better than 2.x to get those package distributions ported, it's gotta be substantially better. One way to make it substantially better is to get the new [Distribute](http://pypi.python.org/pypi/distribute http://pypi.python.org/pypi/distribute) finished and fast-tracked into 3.x (Distribute is a replacement for setuptools).
- masklinn 17y ago
  The first important step is the rewrite/refactoring/improvement of distutils, which is going on right now.
  - garnet7 17y ago
    Well, I hope that whichever Yaks need to be shaved in the process (PEPs written, approved, implemented, whatever) get shaved expeditiously so things can move forward.
rbanffy 17y ago
The article also indicates we seldom use about 99.04% of all packages available. It's a believable number
tvon 17y ago
Well, the topic in #python includes "It's too early to use Python 3.x".

16 points by illumen 17y ago

It's often said that Java has been held back (feature-wise) by Sun's obsession with downward compatibility. This article indicates that it may be preferable to the alternative... What's the point of new features when hardly anyone uses them because not only their own code but also most of the important libraries don't support the new version?

chaosmachine 17y ago
Drupal suffers from this same problem. Every time a new major version comes out, all of the existing modules break and have to be rewritten. With over 4000 modules (and roughly the same number of module authors), it's a serious problem. What ends up happening is everyone keeps using the old version of Drupal until they're forced to upgrade or abandon it due to security concerns (old versions don't get patches from the security team).
steiger 17y ago
It's of course a trade-off. I think the changes that made into Python 3.0 were welcome, and breaking downward compatibility was a hard but acceptable consequence. It's a difficult matter: Let the language evolve and break things, or make it grow old and stale? Careful ponderation should be made.
amix 17y ago
I don't think backwards compatibility is the problem. The problem with Py3k is that it does not offer any vital features, it requires a lot work to port to Py3k (to top that off Py3k is slower than Python 2.x). The only way I can see Py3k successful is if there are clear advantages to move to Py3k, like improved performance or extra vital features (like improvement of inline functions (lambda)). The improved performance is also a big question as teams like PyPy and Unladen Swallow are working on improving Python 2.6 and not Python3k.
- eru 17y ago
  > extra vital features (like improvement of inline functions (lambda)) I would not consider this a vital feature. You can already nest arbitrary functions, you just have to name them. However Tail-Call-Optimization would be a vital feature. (Alas, it's not going to happen.)
  - masklinn 17y ago
    I'm pretty sure the improvement he was talking about is "full" anonymous functions (w/ multiple statements & everything).
    - scott_s 17y ago
      And I'm pretty sure eru understood that, which is why he pointed out you can achieve the same functionality by naming them.
      - eru 17y ago
        Yes. And if you want multiple statements, you are either doing site-effects (evil) or using variables inside your closure, so you might as well name it, too.
        masklinn 17y ago
        > you are either doing site-effects (evil) Uh not necessarily, and even if you are there are many situations where it's interesting and useful. > or using variables inside your closure, so you might as well name it, too Does not follow.
- jnoller 17y ago
  How about this one: 2.x is dead. While bug fixes and minor changes are occuring, it's going to slip into maintenance mode once 2.7 comes out. All the cool things, such as importlib and other items are py3k only.
masklinn 17y ago
> This article indicates that it may be preferable to the alternative... No. Python 3.0 was never expected to see widespread adoption, nor was it expected that libraries be ported to it. Ever. Python 3.1 is expected to have a better adoption rate, and library porting to start in earnest during its time and even more with Python 3.2 next year, especially with tools such as 3to2 appearing. Everything is going according to the plan, and this plan is the reason why 3.1 was released less than 7 months after 3.0 (3.2 will probably be released early next year, putting barely a year between 3.0 and 3.2). That's also why 2.7 is in the works (and was already planned before the 2.6/3.0 releases), and there might even be a 2.8 (though that's unlikely). edit: when you get down to it, those claiming Python 3 has failed are either morons or FUD-spreader, Fox-style. I'm not saying Python 3 has succeeded, but we won't know whether or not it has failed before the end of 2010 at the earliest. Right now, everything's going according to the plan. Please remember the planning for Python 3000 started for real during the 2003 State of the Python Union. Meanwhile the release of Python 3.0 was in december 2008.
- jnoller 17y ago
  I love you
  - masklinn 17y ago
    I love you Well I don't know if I love you but you sound like somebody I could love.

garnet7 17y ago

Python 3 can't just be better than 2.x to get those package distributions ported, it's gotta be substantially better. One way to make it substantially better is to get the new [Distribute](http://pypi.python.org/pypi/distribute http://pypi.python.org/pypi/distribute) finished and fast-tracked into 3.x (Distribute is a replacement for setuptools).

masklinn 17y ago
The first important step is the rewrite/refactoring/improvement of distutils, which is going on right now.
- garnet7 17y ago
  Well, I hope that whichever Yaks need to be shaved in the process (PEPs written, approved, implemented, whatever) get shaved expeditiously so things can move forward.

rbanffy 17y ago

The article also indicates we seldom use about 99.04% of all packages available. It's a believable number

tvon 17y ago

Well, the topic in #python includes "It's too early to use Python 3.x".