On Mer 20 avril 2005 11:04, Philipp Lohmann - Sun Germany a écrit :
Face it: the patch acceptance is not that bad, It's just there are not many patches.
Quoting (from memory) a developper on a Linux list (answering why a feature had reverted to a suboptimal state after being fixed during a few months)
"I fixed this when working on foo problems last year and it was accepted in the Ximian branch, but not the Sun one. After spending a year rebasing the patch for all releases I've decided to drop it and work on bar instead."
There certainly were some problems, but many of the causes are known and are being worked on.
One issue with patches that appeared on the Ximian branch but not on OOo was a disagreement about what may or may not be put onto a released, stable branch.
The OpenOffice.org project is rather restrictive about what is accepted on stable branches. Basically only bug fixes, not features are accepted there. And even for bug fixes the risk is weighed against the benefit. Few, if any, UI changes are accepted, because that would mean documentation and localization would have to be respun as well. Incompatible changes that break (internal) interfaces are not accepted there. There are several reasons for those policies.
Ximian was eager to move forward with new features. They maintain a OOo 1.1.x stable + new features version. So they more readily accept feature and UI patches against that branch. They even backported several features created on the development branch by Sun to the 1.1.x codeline. Additionally they appear to care less about binary compatibility (only caring to be compatible within one release of their distro). They are focusing on Linux exclusively, so they can accept patches that work only for Linux or Unix.
Many patches were accepted or would have been accepted by OOo on the development branch. Most work done by Sun developers happened on that branch. Many patches against the stable branch did not meet our criteria of 'stable', so it would have been necessary to forward port them to the development branch. This includes a lot of what went into Ximian-OOo. The development branch is/was a quickly moving target, so it was not always easy to follow. With the advent of the child workspace (CWS) model, this should have become much easier - and the number of changes contributed by non-Sun developers with CVS access has risen significantly since (my personal impression, I have no numbers).
One of the reasons why contributors did not feel satisfied with patch acceptance only on the development branch is the long release cycle. If there are 18+ months between stable releases and feature stop is 6 months before the planned release date, it takes a very long time for a contributed patch to appear in an official release. The regular release of milestone builds was a first step toward mitigating this problems. For the future there is an ongoing discussion (started in public at the OOo Conference last fall) how we can get a shorter release cycle, which would address this problem (and others as well). The CWS model ought to make it possible to keep the weekly builds near release quality, which would be another step towards eliminating this problem.
And last but not least, you can always, for any OSS project, find cases where the project lead rejected patches, because they did not fit in with project policies or roadmap. Citing examples won't help this discussion much - we can find many cases of accepted patches as well.
Ciao, Joerg
-- Joerg Barfurth Sun Microsystems - Desktop - Hamburg >>>>>>>>>>>>>>>>>> using std::disclaimer <<<<<<<<<<<<<<<<<<<<<<< Software Engineer [EMAIL PROTECTED] OpenOffice.org Configuration http://util.openoffice.org
--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
