[Qgis-developer] Help us plan for QGIS 3.0

[Richard Duivenvoorde]

On 17-01-16 23:38, Nyall Dawson wrote:
> On 18 January 2016 at 09:03, Tim Sutton <tim at qgis.org> wrote:
>>
>> Dear QGIS Developers
>>
>> For some time now we have been talking about moving to 3.0. The PSC is looking for proposals on how to manage the process of moving to QGIS 3.0. For a little more context please see the blog post I have made [1]. Once we have all the proposals in (please see the example from Matthias Kuhn at the bottom of the aforementioned post and keep it nice & simple)  we will collate them and then select one (or a hybrid of several) and share those plans with the broader community. I will make all the proposals publicly available and we will make the final decision in consultation with the developer community.
>
> Great summary Tim! Thanks for helping push this along.
> 
> It's probably no surprise (since Matthias and I have spoken at length
> regarding this), but I'm in favour of Proposal 1. I just don't believe
> we have the resources to support any form of parallel development like
> proposal 2 requires.

Hi Nyall,

I also agree with your argument that we do not have resources to support
parallel development.

But I do not agree with the two plans Tim proposed.

I would favour the project to concentrate fully on a QGIS 3.0. I thought
we already spoke about this one year ago: 2.14 would be the last LTR in
the 2.x serie so current 2.x users can at least use that one (plus ltr
fixes if really needed).

And the project could then fully focus on a good Qt4, Py3, new api version.
Whether this is in master branch or in a separate qgis3 branch (so we
can fall back if really needed) I am not sure.

I think without full focus on a transition it just will not happen.
Personally I think current options are too much of a compromise, and we
will end up with a lot of wrappers and hacks to keep 2.x alive while
also trying to server 3.0...

But I'm not a core dev so devs feel free to convince me that *for the
project* it is better to mix two python/qt/api versions and try to hide
that (as I see it). At the moment I just do not believe in this approach.

Regards,

Richard Duivenvoorde