On Jan 13, 2013, at 9:11 AM, Benny Malengier <benny.malengier@gmail.com> wrote:

2013/1/13 Paul Franklin <pf.98052@gmail.com>
On 1/12/13, John Ralls <jralls@ceridwen.us> wrote:

> We don't need Py3 to release. Py2 is available everywhere.  Don't worry
> about it. The few people who care can get the py3 forks from the respective
> repos and roll their own.

We might consider making Python 2.7 the mandatory one and Python 3
the "optional" one, perhaps mentioning in the README that Python 3
paths are less likely to have been thoroughly tested.

I know that Benny has put an immense amount of work into making
it all work on Python 3, and will probably be very disappointed, but
it also seems to me that changes are being discovered all the time
(division changes, string/unicode changes, encoding changes, etc.),
every time somebody tries a code path under Python3 which has not
yet been tried.

Perhaps this would make it easier to have it on Windows also, I am
not sure.  But I don't think we should forget our Windows users, since
I am guessing they are the majority.

Although you are right, packagers often don't want every year to redesign their spec files or build system. As they will have a lot of work for gramps 4.0, doing it immediately for python3 and being good for the coming years, has advantages for them.
Nevertheless, for packagers, it should be easy to create a gramps in python3 and a gramps2 in python2. It is up to them what they actually do.

Main point as I see it is that in the very near future, developers will no longer test python2 code path, as in 2013 most distributions will have python pointing to python3. So although python2 is the safe bet at the moment, this will change very fast. As the original idea to have 4.0 unstable only is no longer going through, I don't mind indicating in the README that python2 will normally be the more stable option for 4.0

About windows, as Helge shows, the system there is not ready for 4.0, which is the reason 3.4 will be supported longer. Programs like Gramps moving forward is exactly what is needed to fix the issues on Windows.

Nope. Not happening,. GObject-Introspection doesn't work with python3 [1].  Neither does libxml2 or libxslt, so Gtk-doc doesn't, either.

Just out of curiosity, look through Ubuntu admin scripts written in python and see if they've all been ported. 

You seem to be under the impression [2] that most distributions are going to be able to support running Gramps4. That's absolutely not the case. As of today, only the bleeding-edge distributions like Ubuntu and Arch can run Gramps 4 without the user having to build a Gtk stack from source. It appears that Fedora 18, to be released Tuesday, will. Debian unstable (Sid), which won't be released as stable for 2 years, is still on pygobject 3.2.2. RHEL 6 doesn't even support Gtk3 at all, though that costs money and it's likely that few Gramps users will pay for it. OpenSuse is currently on 3.2.2, though they have an RPM for 3.4.2 in their repo that might make it into 12.3 in March; otherwise it has to wait for 13 in November. Aunt Martha isn't running Arch. If she isn't on M$Win, she's probably on Fedora 16 or Suse 11.

John Ralls

[1] https://bugzilla.gnome.org/show_bug.cgi?id=679438
[2] http://www.gramps-project.org/bugs/view.php?id=6288#c26965