It is highly recommended to remove Python 3.3 support in order to reduce the package size, and further activate the plug-in market enthusiasm!
It is highly recommended to remove Python 3.3
jfcherng
#3
I don’t have statistics but I bet that >90% plugin still runs on py33. By making py38 optional, we save more space, since py38 size is bigger than py33’s!
OK I am joking. Staff knows that but it just won’t be done in a foreseeable future… Plus, it’s highly recommended to remove py38 (EOL in October 2024, the last version officially supports Windows 7).
But I also remember that one of the staff (probably wbond?) said that “we are totally capable of maintaining py33 even if it’s EOL, so it’s fine”.
0 Likes
Gfcaer
#4
There are two Python engines, it feels weird, it doesn’t matter which one you get rid of:grinning:
0 Likes
Gfcaer
#5
Disadvantages of two Python engine: one is to occupy a large disk space; Second, it takes up a lot of Internal memory, and I feel that most plug-ins can add a “.python-version” file for PY8.
0 Likes
jfcherng
#6
Agree to remove py33 but honestly I don’t think so. What takes up a lot of memory are plugins, not the plugin host. And I won’t call 10MB a large disk space either.
0 Likes
deathaxe
#7
The recommended plugin environment for new plugins is python 3.8. New packages must target it to be accepted in Package Control’s default channel.
Python 3.3 plugin_host just exists for backward compatibility with existing 5K packages out in the wild.
As most of them are rather unmaintained, chances are low for all to be migrated to 3.8. So preconditions to remove python 3.3 are just not met. A tiny HDD memory saving of few MBs while most SSDs provide TBs of storage just doesn’t justify breaking compatibility with the huge amount of available plugins.
Most of those plugins may run on python 3.8 just fine, but this can’t be assumed in general. Thus each plugin must actively opt-in to python 3.8, which requires their authors to take action.
Package Control 4 supports opting-in packages via channel/repository meta data, but unless packagecontrol.io supports the required repository schame, this is no option either. Even after that, someone would need to review each of the 5K packages to check compatibility, manually.
Once someone found the passion to do so, we can re-do the job for python 3.20
1 Like
Gfcaer
#8
thanks.
It is understood that the current activity of most plug-ins is not enough, take this opportunity to remove the support of PY3.3, to a certain extent, can stimulate the activity of the plug-in community, let everyone to re-understand ST4
0 Likes