Re: Proposed external dependency: WebKit/GTK+

["Alp Toker" <alp atoker com>]

2008/11/6 Willie Walker <William Walker sun com>:
> For me, one of the most important problems we face with WebKit is its
> lack of accessibility support right now.  I have a great fear that Alp
> may have grossly underestimated the scope of the work.  I have some
> confidence, however, that the WebKit folks who were at GNOME Boston
> should be able to do a decent analysis of what it will take to add it.

Hi Willie,

I'm not sure what you think is being underestimated here. Instead of
throwing our hands in the air and despairing at the size of the task,
my company (Nuanti Ltd.) and I made the decision to make WebKit GTK+
accessibility happen early in 2008. Since then, we've been working
hard to develop formal accessibility via ATK as well as improving
keyboard navigation capabilities. As of November 2008, nobody else has
contributed a line of code and David Bolter is the only external
contributor to have even CC'd themselves on any of our WebKit GTK+
accessibility bugs. There have been no responses in those bugs to
requests for testing of the patches.

Two months ago, Nuanti Ltd. which does consultancy and development on
WebKit even introduced a reduced rate where we set aside one developer
day a week to accessibility on any WebKit-related project. One company
we're working with has already accepted this sponsorship arrangement.

We also have a formal testing strategy for WebKit GTK+ now which
involves assessing the actual end-user accessibility of smaller
non-browser applications like Devhelp and Yelp.

Meanwhile, the latest ATK patchset we posted to WebKit bug #21546 also
adds initial support for certain DOM events and ARIA accessibility
enhancements. This work is not just being done to "tick the box".

It's only now that you're mentioning you tested this patch and it
didn't work for you. Clearly the functionality we're developing _does_
work for us both in the testing environment and with real-world use --
if it didn't work for you then, I don't understand why on earth you
didn't take the time to mention it to the developers on the bug
tracker so we could have had your issues fixed by now. Even direct
personal email to the developers listed in the ChangeLog entries would
be better than your complete silence that has so far been broken only
when the perennial question about why the WebKit transition isn't
complete yet gets raised on this mailing list.

>
> I so wish I wrote down all the names of the WebKit folks that were at
> GNOME Boston.  I'd like to follow up to see how their exploration is
> coming along.

It's not a question of exploration at this point but continued
development to cover any missing functionality that will make the web
accessible in a full-blown browser. We can make this happen but we
need constructive input from developers and testers.

As to your concern that this is an incredibly difficult and possibly
unattainable task, I'm going to paraphrase from one of the bug reports
that you aren't subscribed to:

1) The WebKit GTK+ accessibility layer reuses much of the established
code that is deployed in WebKit/Mac (Safari). This greatly reduces the
total amount of code we need to maintain and reduces the problem down
to simply observing and adapting to the differences between their a11y
SPI and ours in many places.

2) When developing the WebKit GTK+ ATK implementations we studied what
was in Mozilla and GtkHTML. Mozilla's accessibility layer contains an
_enormous_ amount of redundant boilerplate which we've developed some
simple strategies to avoid. This is the exact same advantage that has
let us develop a first-class Cairo graphics backend and Pango-based
international text support without the vast resources and time it's
taken in other browser engines. Our code is direct GTK+, ATK etc. and
we reap the benefits of reduced complexity from this every day.

The basics are already here and first class web accessibility is
around the corner. It will happen sooner if we work together.

I'm also very grateful to the organisation that's sponsored our extra
work on the accessibility layer, even if it was just a few days on top
of what we're already putting in. This kind of sponsorship (and indeed
direct targeted projects on accessibility) helps the effort as a whole
and has given a great morale boost in knowing that the world cares
about WebKit GTK+ accessibility.

Let me know if you will have the time to study the current state of
the ATK and caret navigation patches (the ones in the bug tracker are
a couple of weeks old) and we can start working together today and get
this synced up to WebKit SVN within the week!

Cheers