[Qgis-developer] Time for 1.0 API freeze

[Tim Sutton]

Hi

2008/11/20 Martin Dobias <wonder.sk at gmail.com>:
> On Wed, Nov 19, 2008 at 10:28 PM, Tim Sutton <tim at linfiniti.com> wrote:
>>> > From the PSC meeting we will temporarily unfreeze trunk to allow
>>> > improved labelling to be added, so I will hold on for API freeze until
>>> > after that - feel free to tweak on if you spot inconistencies in the
>>> > mean time!
>>>
>>> When will this happen the inclusion happen? isn't improved labeling
>>> scheduled for 1.1?
>>
>> It was but the feeling at the PSC meeting was that we should unfreeze
>> trunk for labelling (not a general unfreeze) and get label collision
>> detection in before we API freeze and release QGIS 1.0. Still to be
>> decided who / how we will implement it. I was wondering if we could use
>> graphics scene collision detection. Do they still have that or was it
>> lost in move from canvas? We are still hoping for a december release
>> though not sure if that will be possible...
>
> Well... there's practical problem - who will implement it :-) It's not
> something that one can develop in two days. There are more problems I
> see:
> - the release will be shifted again and again. Real release date can
> be moved to March or so
> - long feature freeze is not good
> - we're parting the original idea of frequent releases
> I understand that good labeling is quite important but if there isn't
> any code, waiting for full (and bug-free) implementation can take
> ages. Please re-think this!
>

To make it clear this wasnt my sole personal decision, but a decision
of the PSC. Personally it think whatever we do we are going to make
people unhappy since there are at least the following trains for
thought:

- release now its been too long since 0.11
- release now but postpone API freeze
- dont release now there are too many bugs
etc. etc.

In reality I agree with your statements to some level and would be
more happy to release QGIS 1.0 with stable API in December. However I
also see some merit in the other points of view. You should take this
up on the PSC mailing list suggesting a concrete course of action.


> And there are even more reasons to release soon - we're on halfway
> between 0.11 and 1.0. There's evolving community around python
> plugins, people are trying plugins or developing new ones. You can see
> that without proper 1.0 release there will be quite some problems with
> incompatibility - like nowadays, every few days someone posts a mail
> about plugins that are not working because of API changes. 1.0 release
> will have checks for incompatible plugins so this problem will go away
> quickly.


Once again I agree these are good motivations for a December release.
The discussion on PSC was that the labelling freeze would be lifted
but no extension to release timeline was planned. As you say this
likely means even if labelling does make it into svn by Dec it will be
not well tested. We also need to get clarity on who will implement it
and by when otherwise as you say, we will be sitting in limbo. Lets
take this onto the PSC list and we can discuss further and get some
clarity as to how things will proceed...

>
> Regarding your question about graphics scene - yes there is collision
> detection. Qt uses BSP tree indexing. But I'm not convinced that it
> has be used for labeling. We can use spatial indexing based on R-tree
> included in qgis core library. I don't know which one is faster for
> this kind of use.

Ok thanks for the info.

>
> Martin
>



-- 
Tim Sutton - QGIS Project Steering Committee Member (Release  Manager)
==============================================
Visit http://linfiniti.com to find out about:
 * QGIS programming services
 * Mapserver and PostGIS based hosting plans
 * FOSS Consulting Services
Skype: timlinux Irc: timlinux on #qgis at freenode.net
==============================================