Going forward

[Boudewijn Rempt]

Hi,

Krita defined as a application that offers for developers a good platform for 
experimentation and for users unique and surprising features (see my other 
mail) needs to have a stable base to build on. So in this mail I want to set 
out what I see as the way forward for us, _if_ we conclude that going forward 
with Krita in the current way is useful. If not, we need to think again.

I'm assuming KOffice will go for a 3-month feature release scheme where 
features will be integrated when they are ready, so not destabilizing the 
apps.

For the 2.1 release we are focussing on stability a lot, and I think that's 
already really paying off. At least, we keep about the same number of crash 
reports, but it's different crash reports every day, so we're fixing bugs.

For 2.2, I think we should really go for performance, profiling, analyzing, 
fixing -- even if that makes a refactoring necessary. The following areas have 
known issues:

* tile engine (the old one has the BKL, the new one doesn't swap yet & isn't 
yet really fast)
* mask generation for brushes (we analyzed this in Berlin)
* iterator caching
* projection recomposition: we recompose everything immediately, even pixels 
that are not visible, so region-of-interest might be a good first step.

I'm sure we'll find more performance issues once we get down to it. This is 
also a good moment to fix our api's and document our application architecture.

Then, for 2.3 and after, I hope we have our stable and fast platform and can 
set out adding the things we love, new brush engines, cooler user interface -- 
whatever, maybe even focus on users' workflows?

-- 
Boudewijn Rempt | http://www.valdyas.org