I started writing this in a comment on the Firefox thread, that's still going on and in interesting directions, and realized it should be in a blog post.
The point was made that Mozilla was headed in a good direction with their support for Firefox plug-ins. But now, by breaking them, as they bump the version number more frequently, they're turning in a less optimal direction.
In 2008, I stripped the extraneous stuff from the OPML Editor to get something that was just an app-running environment and an outliner. This was a good move. I had added a layer that I never finished and didn't even like, and it took two years to fully realize that, and take the time to clean it out and get a fresh start.
At the same time I added a Tool Catalog page embedded in the app that makes it one click, with confirmation, to install a new app. Another move that has stood the test of time. And one that, theoretically, removed any reason to bundle user-level functionality.
And then earlier this year, I did it again -- and added Blork and River2 to EC2 for Poets. Neither was mature enough for that. I reached too far before it was time. Again. Anyway, now I'm creating a stripped-down EC2 for Poets. It'll be a little more complicated to set up Blork, but a lot less complicated if you want to use it for something other than Blork. Maybe, at sometime in the future when Blork is fully mature and all the roughness has been smoothed out, it can hide in the corner of the install, and not get in the way if you're not interested. But we're definitely not there yet. And it was in the way of something I wanted to do. A sure sign I had reached too far.
Something that works well, is adding api-level support for web services. For example over time, the glue that connects up to various Amazon services has gotten really rich. Just added support for Route53. If an app doesn't want to use it, no problem. The cool thing is that all those services are configured by entering two strings into a dialog. Not too bad.
I guess if I were evolving Mozilla, I'd try to fork into two threads. One that's stable, and only gets upgrades to the rendering engine and addresses security issues. A totally stable plug-in-running platform. And the upgrades are coordinated with developers, which includes people who manage large networks of installations of Firefox users. If Mozilla doesn't want to do this, they could license another company to do it. Look at how HP just reorganized to rebuild itself around software. There's value here, and it's a shame to see it thrown away.
A really bold move here would be to do a partnership with Microsoft, at a time when the open source world has a lot of leverage with them. Think about the enemy of your enemy being a friend. Think of a mature company that can set up and run a vast service network for enterprise software. A developer conference on the scale of WWDC or Google I/O.
The second fork would be the current rock and roll thread of development. By splitting in two you'd free yourself of the drag that you must be feeling and believe me, it won't go away unless the people really go away -- to Chrome.