Re: [gentoo-dev] unsanctioned python 2.7 crusade

[David Seifert]

On Thu, 2019-12-05 at 21:56 +0100, Thomas Deutschmann wrote:
> On 2019-12-05 21:31, David Seifert wrote:
> > > On another topic, I'd prefer for python 2.7 not to be removed
> > > from
> > > gentoo. Tons of code still uses it.
> > >
> > Sorry, but I'll have to disagree with you on this.
> >
> > We're removing Java too from Gentoo (more implicitly than
> > explicitly),
> > because the Maven/Gradle ecosystem doesn't seem to scale. There's
> > tons
> > of code that uses java and java binaries too, and yet we're
> > removing
> > it. Python 2 is EOL in a few weeks. We have also removed Qt4 and
> > lost a
> > number of useful applications with it. At some point, we're not
> > going
> > to maintain a dead interpreter anymore.
>
> For the records: Nobody in this discussion or IRC chat said
>
> > Keep Python 2 forever.

Again, disagree. You'll hear lots of voices that are along the lines of

  "So much enterprise code won't get ported to py3, and RedHat will be
  maintaining RHEL 7 and 8 for the next 10 years, so we'll always have
  security patches to rely on. Let's just keep Python 2 for the
  foreseeable future."

many Gentoo devs have voiced that opinion, so asserting that noone says
"Keep Python 2 forever" is false, and not by a negligible margin.

>
> It's about timing. From my POV and I read
>
> > Tons of code still uses it.
>                ^^^^^
> the same,  there is no need to mask any Python 2 stuff _today_.

When we started removing Qt4, tons of code still used it. To put things
in perspective:

grep -rl 'IUSE.*python_targets_python2_7' /usr/portage/metadata/md5-
cache/ | wc -l

gives me 7070 ebuilds currently. 7070 is easily more than one and
closer to two orders of magnitude more ebuilds using python 2 than Qt4
back in the days. Removing python2 will turn into a multi-year,
monumental effort of epic proportions, and I'm willing to bet
€1000 that we'll still be stuck with it in 3 years. It will be one of
the largest undertakings of Gentoo, probably more involved than getting
rid of EAPI=5 ebuilds. Removing maintainer-needed and other semi-dead
packages is part of a proactive strategy in continuously removing and
treecleaning stale stuff from the tree. Tons of java stuff also still
works, yet we're removing it because the ebuilds are ancient and
unmaintained.

>
> Especially when new Python project lead sent a mail [1] to this list
> few
> weeks ago stating that there will be a _new_ last Python 2 release in
> April 2020 (120 days away!).
>
> Now please explain to me and any Gentoo user depending on Py2-only
> software why you are taking actions 120(!) days in advance.
>
> Again, nobody wants to keep Python 2 forever but starting to kill
> *working* software 120 days in *advance* deserves at least an honest
> justification.

So what do you propose? Starting to remove/fix 7070 ebuilds after April
2020 then? Why start in April 2020? Let's just wait till May 2029, when
RHEL 8 goes into end of maintenance support. That's a good time point
then? It doesn't matter what time point you think is suitable, *any*
time point will be arbitrary to someone. Every change breaks somebody's
workflow.

>
> PS: And given that a release in April won't break the next day, we
> are
> actually talking about more than 120 days in advance. Security
> argument
> is not valid because if there will be any serious vulnerability in
> Py2
> found after this release (which will be surprising after so many
> years)
> you can expect backports because other distributions still have to
> support Py2 two more years at minimum.

And that's exactly the straw-man argument I've been making. You can
always come up with an excuse to delay action on python 2, because
"someone, somewhere, will maintain it". Heck, if RHEL 8 abandons python
2 in 2029, let's just swap in Tauthon then, then we can use python 2
packages till 2100!

>
>
> [1]
> https://archives.gentoo.org/gentoo-dev/message/d00a956180ab7df980ac5642e3abc179
>
>