Python 3 Support: A Plan of Action

Matthew D. Fuller fullermd at
Fri Sep 11 18:51:45 UTC 2015

On Fri, Sep 11, 2015 at 12:10:31PM -0600 I heard the voice of
Richard Wilbur, and lo! it spake thus:
> I haven't kept abreast of the FreeBSD release schedule.  How widely
> adopted are the releases you speak of that include python 2.7.10 and
> 3.4.x?  In other words, we can expect prompt adoption of new python
> releases in new releases of FreeBSD (that's wonderful--on what
> schedule?), but who do we leave behind if we drop python 2.6
> support?

Things operate somewhat differently in the BSD world vs. Linux.
There's a separation between the base system (kernel, ls, ssh, etc)
which is basically a skeletal running system, and the ports tree which
contains all the zillions of software packages you use to make useful
systems (X11, Mozilla, Apache, etc, and topically, Python).

Releases are of the base system; they include ports/packages in the
ISO's, but they're just a snapshot of the point in time.  Users either
use the ports tree (basically a large skeleton of recipes for building
packages, like Debian packaging recipes, except implemeted in a
fever-nightmare of make(1) for hysterical raisins) to build/install
software, or use pre-built binary packages which are built from ports.
And ports isn't branched[0], so all (supported, naturally) releases
use the same set of stuff.

So anybody who's keeping up to date[0] will have whatever's in ports
now, give or take a couple weeks or maybe a month, depending on how
often they update their systems.  Of course there are always outliers
who might just not update their systems for various reasons, but
they're of less concern by virtue of being outliers.  And as plain bzr
users, they'd be using the packaged bzr that already matches their
other stuff anyway.  Unless bzr starts improving and changing at crazy
speed (which wouldn't be ALL bad, but certainly doesn't seem likely)
that's not a meaningful handicap.

So the short version is, anybody running a supported release and
keeping their stuff up to date is not gonna be more than a couple
months behind the current head state.  People outside those boundaries
are far enough out on the fringe I'd ignore them.  At any rate,
whatever CentOS/RHEL we have to care about is gonna set the bar way
further back.

Nobody[1] is gonna have py2.6 around unless they haven't updated
things in multiple years.  There are probably orders of magnitude more
people with 3.3 than 2.6, but even they are at this point a bit behind
and either sticking with that for local reasons, or had it installed
[from ports; pkg users will have moved when the default changed] prior
and haven't bothered changing.

[0] Technically not true anymore; there are now quarterly branches, so
maybe in some case we care about <=3 month old changes.  But for
practical purposes, close enough.

[1] Well, there's probably SOMEbody for quirky local reasons
maintaining it on their system, but for our purposes, they don't

Matthew Fuller     (MF4839)   |  fullermd at
Systems/Network Administrator |
           On the Internet, nobody can hear you scream.

More information about the bazaar mailing list