"I think the Firebird development community is too small and too exclusive and thus tends to change things whose original rationale they do not know." - David Baron
Indeed. There is a heated discussion on the mozdev project owners mailing list about this topic lately. Developers are getting frustrated by the direction Mozilla is taking and are unsure about developing for a moving target.
Mozilla Firebird, and the focus of the Mozilla Foundation on end users has only worsened this issue. Mozilla Firebird has had very little focus as a developer product for building applications. Sure, the extensions are there, but they
target the Firebird UI directly, and since Firebird is a 0.6 release, the developers accept the ongoing changes to the base product.
The Seamonkey framework hasn't changed that much lately. Changes to XUL have been generally insignificant since 1.0. Firebird and the new toolkit seems poised to change that however. Sure, adding features such as custom toolbars are great, but why the change from the base XBL files being in /bindings to now being in /widgets? Compatibility seems to have been intentionally broken.
I think that the Firebird developers began to focus too much on the
'simple product' user and lost track of the developer.
With regards to API changes, I think freezing interfaces that have some flaws now is better than rearchitecting them, breaking compatibility, and then freezing them. No matter how you improve the interfaces, there will always be something wrong with them.
One problem though is that Mozilla is significantly more customizable and extensible than other products. Many extensions actually rely on specific elements of the UI or other aspects to remain constant. If your extension
depends on there being a Back button with a menu that works a certain way,
of course it's going to break if an improvement is made to the base browser product.
Added to which are the hordes of people with nightly builds, beta releases, and people with the Mozilla suite and others with Firebird, and you end up with
a vast array of incompatibility problems to deal with. Remember though, mozilla.org has only ever in their history released two API stable browser products. The first was Mozilla 1.0 and the second was Mozilla 1.4. Remember that it's only the hordes of users that want it to work on every release, not the Mozilla Foundation. I don't think that nightly builds should have any requirement to be compatible entirely.
The original point dbaron made was that developers are changing things that they don't know much about. Over time, this usually happens. The reasons why something was done a particular way are lost over time. Specifically, he was referring to the way that the font settings in Firebird were moved deeper in the preferences.
As I said when work on Firebird was just starting (then called Phoenix), the goal of the Firebird project is to make the browser simpler, not to make it more usable.