Gecko Is A Losing Bet

Gecko is behind Webkit on Web features, performance, code footprint, and
code quality. Web features where they're ahead include HTML5 SQL, HTML5
drag-drop, downloadable fonts, CSS transforms and animation, CSS media
queries, CSS box-shadow, text-shadow, text-stroke, text-overflow, CSS
display:compact and run-in, CSS3 borders and backgrounds, SVG <img>s,
two-pass layout and frame flattening for mobile (see Android).
Performance: almost every third-party JS/DOM and pageload perf test
shows us behind, sometimes way behind; we can quibble about individual
tests but the trend is clear. Code footprint: I don't think there's any
argument. Code quality: everyone who's hacked on both prefers Webkit.
Gecko has many areas that need major rework or are unowned and rotting:
e.g. HTML parsing, docshell/history/bfcache, focus, nsIFrame
continuation model (fundamentally broken), abs-pos containers,
selection, editor, floats, frame construction, inline layout, XUL layout
especially box/block interaction, logic in frames that should be in
content, plugins, imglib, printing, view system (to be removed),
XPConnect, DOM/native event split, embedding APIs. Not to say Webkit is
perfect, but we've got a lot of work just to prevent gangrene.

Where are we ahead? In Web features: JS1.7 and E4X, a few DOM APIs, some
text rendering and I18N, accessibility. We seem to have a better data
footprint for page loads. We talk about Web compatibility but I think
that's almost entirely a market share issue, and Webkit is getting there
and surely will have "enough" with GBrowser.

Of course we can fix everything where we're behind eventually, but in
the meantime they're not standing still. Will we even progress faster
than them (Apple, Google, the Qt guys and a growing open community, with
a cleaner base) and be catching up over time? I think not.

I used to think that if Firefox gets crushed by Webkit-based browsers
and drags down Mozilla, that would be sad and we'd go and work for Apple
and Google and c'est la vie. But Brendan and Mitchell convinced me
that's not an acceptable outcome, so I believe betting Mozilla on Gecko
is a big mistake and we must change course.

A Better Bet

Suppose we were able to get Firefox and the rest of the XUL ecosystem
running on Webkit. This would facilitate Mozilla's survival, by removing
a key competitive edge that other browsers have over Firefox and letting
us focus on our strengths --- brand, mission, user community, open
source community, extension community, XUL app community. But it would
also be tremendously good for the Web. Our combined market share would
be a stronger lever than ever against IE-only holdouts. Web developers
would be happier with one less browser engine to target. Embedders
wouldn't have to make difficult choices. But most importantly, with the
open-source browser community and its backers pulling together on a
single code base we would make progress so much faster against IE and
the Web competitors.

It's true that two open-source engines gives some benefit in standards
orgs, but there would still be multiple stakeholders with their Webkit
variants, so I think we'd be OK there. Ditto for the benefits of
competition. In the long term, a hegemony of Webkit derivatives could be
a reasonable sweet spot as the implementation of the open Web, IMHO.

Getting There

We'd have to port XUL and XPCOM infrastructure over (including XBL,
probably as XBL2) and fill in other bits of Webkit that are broken or
missing. But however hard that problem is, it's not getting bigger over
time (unlike the effort required for Gecko to catch up to Webkit...). In
fact it gets smaller as Webkit gets more features and bug fixes.

Some of our work, such bug fixes, XUL box layout and XBL2, would
hopefully be accepted into Webkit proper. If Apple hates us and won't
take our patches, it won't look good in their open source community (or
to their own Webkit people). We can use distributed source control to
share with other Webkit stakeholders outside of webkit.org SVN, anyway.

Of course we'd have to keep Gecko moving while working on this. There
are a number of projects that we could do that would add value to Gecko
*and* future XUL-Webkit: for example, a GL cairo backend,
cross-process-shareable profiles+Necko, out-of-process plugins, and of
course, making XPCOM/XPConnect not suck. There's also lots of
work-in-progress that would fuel Gecko 1.9.1 and beyond.

We'd have to make sure the XUL community survives the transition. The
good news is that we've always required testing and fixing with each
major release. Also, I think that a Webkit transition plan would counter
the panic that's likely to set in as Webkit gains momentum, especially
when GBrowser surfaces.

This would obviously be organizationally difficult. There's loss of
control, and a drastic course change with risk, pain, fear and
uncertainty. But I believe we face the latter no matter what. Without
them, working for Mozilla just wouldn't be the same! :-)