Geodog

Thanks for the link. The backend code for themes and extensions has been heavily re-written, and this allows for extensions to be properly uninstalled, and for Firefox to be started in 'safe mode' where no extensions are running (in case an extension causes some kind of problem). They are also 'sandboxed' - they can't overwrite program files so they're less likely to conflict or corrupt the installation. There's also now an auto-update function.

The themes and extensions do all need to be repackaged to work with 0.9 - this means that the author of the extension probably needs to spend a few minutes ensuring that the extension files are in the required folder structure in the XPI file, and then including a new metadata file. For themes, it's even easier as only the last step is really required. I imagine that the vast majority of themes and extensions will be 0.9-compatible very soon.

0.9 is a feature-complete release so bar some further improvements to the default theme (which the authors admit is '0.1' quality), there will be no new features before 1.0, just bugfixes. That should mean that all themes and extensions will work in 1.0.

Thanks for taking the time to comment, Neil. While it sounds like the new extension architecture will be an improvement, I'm amazed at how little time the authors of extensions are being given to update their extensions, and how badly they are treated in general. When 0.8 was released, downloads of the extensions were shut off.

For me, extensions like Adblock and Tabbrowser are the whole point of Firebird -- without them there isn't any reason to use it.

I totally agree on this assessment of Firefox and the extensions ... I far prefer Opera right now.

All these conflict/update issues are nowhere to be found in Opera .... it just packs a more powerful punch in a smaller size .... I don't know how they do it, but they've done it.

http://www.sidnak.net/opera.html
http://nontroppo.org/wiki/WhyOpera
http://tntluoma.com/opera/lover/7/

The extension overhaul was for good reason, as is the automatic disabling.

You may remember:
http://www.mozilla.org/products/firefox/releases/fix-extensions.html

That is what a large number of users would be seeing were it not for this new code.

I posted documentation detailing the required changes on May 2:
http://forums.mozillazine.org/viewtopic.php?t=73423&start=0

... approximately six weeks before the release. Given that repackaging an extension that doesn't use unfrozen apis takes approximately 10 minutes, I think that's plenty of warning. Extensions that use unfrozen apis are usually broken by each new release anyway, they use those apis at their own risk. That's not a function of Firefox, that's a function of Mozilla in general.

It'd be nice to give everyone as much time as the needed to update their stuff, but we're dealing with the realities of the market here, windows of relevance, etc. We need a 1.0 this year, and to do that we need an accelerate pace.

Ultimately most of our users won't be using extensions, and those that do expect to be able to do so without having the app fall apart every time they upgrade it.

Ben,

I do remember the XUL binding error, which I had thought was caused by an API change. As I said in response to Neil's comment, it sounds like the new extensions architecture is a significant improvement.

What I don't get is the disregard for those of us who are already Firefox users who use, or even depend on extensions. Requiring new versions of extensions then shutting off easy access to extensions with the release of 0.8 was a pretty strong disincentive to upgrade.

Not taking very much time to test the new extensions architecture seems foolhardy, especially for us heavy extension users. You argue that time to market is your highest imperative, but switching to an immature default theme with 0.9 doesn't seem consistent with that imperative.

As I said above, for me extensions are the best part of Firefox, and the first time I wrote and recommended Firefox (firebird) last year, I did so with a list of recommended extensions. Every review I have seen of Firefox mentions the extensions positively.

I would be curious if you have empirical evidence for your assertion "Ultimately most of our users won't be using extensions", in terms of say Firefox vs. adblock downloads, or if that is just a working assumption? If the latter, it might be worth testing your assumption.

In any case, thanks for taking the time to comment.

Extensions work by overlaying material into the browser. In the course of active browser development, the hooks that certain extensions rely on changes, the functions that they might rely on change too - these are not frozen entities - and likely never will be. Extension authors (myself included) must track these integration points from release to release to make sure things are up to date. Basically Extensions are a really clever hack.

In the future we may expose some standardized integration hooks in order to make extensions behave in a more consistent fashion in a number of ways, but the more ambitious the extension, the more likely it is to tread on something that will change from underneath it in a subsequent release.

In past releases when such things have changed, at best the extension fails and spits a js error into the (hidden) console. At worst it causes a major problem with browser startup such as what Tabbrowser Extensions did for 0.8. There is no safe solution to this other than to disable all of the old extensions. People who suggest that we attempt to load each extension and see if it fails shows a lack of understanding of the way Mozilla and XUL works. Extensions aren't loadable modules with documented integration apis like web page plugins, for example. Extensions can integrate into the app in millions of ways, and there is simply no way to load them beforehand to see if they "fail".

The upshot of all of this was, assuming we touched *nothing* in the Extension system and simply released 0.9 with the Migration engine and other changes, there would be a lot more XBL errors and frustrated users. The only solution is to go in and either remove a specific set of files, hand edit others, etc. Not pretty, and well over the head of our target audience.

The new system will, in the next few months smooth over this process as much as possible by attempting to automatically detect version conflicts, download and install newer, compatible versions. This was no mean feat to engineer and is still not quite ready, but should be by the time it really is needed - the next release.

Firefox is designed to appeal to the widest possible audience, including people for whom computers are not a huge part of their lives - they are not power downloaders, web developers, etc. Extensions are not very interesting to these people. The extensions was provided as a way to let power users of one flavor or another customize their browser to their use case. While the extension system is a major differentiating factor of Firefox, the type of user that uses that functionality will be able to handle the complexity of the upgrade process, especially now that much of it will be automated.

Keep in mind that Firefox is a pre-release product still. We want to make as many of the big changes as possible before we go 1.0 and we get serious business adoption. Fundamental changes are much harder to make after that. I'm not just saying this - I'm speaking from (Mozilla's collective) experience.

Also note that switching to Seamonkey won't get you much - the Extension system it offers has always been far less flexible than Firefox's, and offers none of the upgrade checking that 0.9 has so is basically no stabler than Firefox 0.8 or any earlier version. The only thing that prevents extensions for it from regularly breaking is the relatively slower pace of its UI development due largely in part to it's slow progression towards death.

I think that the necessity of a .91 release made my point sufficiently.

I would still be interested in the empirical evidence for the popularity (or lack thereof) of extensions, instead of your and my assumptions. I think that you are right that a great browsing experience out of the box is the minimum threshold that Firefox's has to cross -- but for enduring success you also need to nurture a whole ecosystem of applications around the browser, as IE has (and as you are doing with extensions).

Thanks again for taking the time to write, and for your part in writing a great browser.