Re: getmail with python 3 adpatations

Am 14.05.20 um 18:55 schrieb Charles Cazabon:
> Matthias Andree <[email protected]> wrote:
>> The thing is that Python 2 has essentially gone out of maintenance with
>> its terminus release 2.7.18 last month,
> [...]
>> Also, if you look at Python 2.7's history, it has seen several massive
>> improvements to its SSL modules in the more recent versions,
> I'm aware of all this, of course.  I strongly encourage all getmail users to
> be using the latest compatible Python releases, though getmail still supports
> older versions of Python for compatibility/support reasons.  There's a lot of
> getmail out there running on appliances that have not necessarily been
> upgraded since they were first deployed.

I understand that a few people will want to run new 3rd party software
(say, getmail) on old 2nd party platforms and come to either of us
upstream maintainers seeking support, but in the end, I get to choose
what I spend my volunteer/spare time on, and that means at some point in
time soon I'll tell people that I no longer care about stuff that's not
compliant to 2011 standards...

Compatibility that comes at little effort, ok, but if it starts with
writing dozens of lines of workarounds for an old system, not gonna
happen, sorry.

Under a stricter maintenance regime, as you explained for getmail
earlier in this thread, it's easier to remain compatible, but then I am
not so sure I should encourage people to mix and match software of many
years apart.

Python release dates with important SSL improvements:

2.7.16 - 2019-03-04
2.7.13 - 2016-12-17
2.7.9 - 2014-12-10

On my software, I've pulled the plug for < 2.7.13 (that's some three
years) for new releases.

> The bit that sticks in my craw is distros/others screaming that the sky is
> falling and they need to immediately drop all Python 2 applications because
> omg omg Python 2 is no longer officially supported.  getmail is extremely
> unlikely to be the source of any security holes just because Python 2 isn't
> getting updated.

Well, this day has been drawing nearer for a few years, was then pushed
back because of the massive amount of Cheese Shop modules not updated
for Python 3.x, but without yet another life extension for Py2.x on the
horizon, distro needed to take it more seriously to jettison (can I use
the word here (not a native speaker) - or relieve themselves of)
technical responsibility and debt - if you ship an EOL package in a new
release, you get to do all the fixes yourself without upstream support.

Distros I am keeping an eye on are still shipping python27 packages, but
I guess those will become fewer over the next releases.

And then again, if people use an old long-term-support distribution,
that's also the point of support for old packaged software and backfixes.

Moving towards Python >= 3.5 or better >= 3.7 (so it doesn't go out of
support in less than half a year's time) would seem the reasonable approach.

Matthias