Re: [orca-list] Moving forward on Firefox

[Jason White <jason jasonjgw net>]

Bill Cox <waywardgeek gmail com> wrote:
> On June 10th, Firefox switched to entirely new code for caret
> navigation and focus control.  I believe this is a significant
> upgrade, but unfortunately, every version of Firefox since then has
> not worked correctly with Orca.  This effects version 3.6 on.

There's a history relevant to this issue. I wasn't involved, so the following
may be very inaccurate.

In response to bugs reported by the Orca developers regarding caret navigation
and focus control, Mozilla concluded that the only way to enable Gecko to
handle this properly itself was to rewrite the relevant code. However, the code in question was sufficiently 
complex and
difficult to maintain that Mozilla developers were reluctant to rewrite it;
hence Orca had to implement its own caret navigation in Python to work around the limitations of Gecko.

In my view, Mozilla caret navigation as implemented in Orca has always been
fragile and prone to bugs. I've reported a number of these, which Joanie fixed
- often through commendable efforts in the early hours of the morning.
Unfortunately, there are still bugs in this area affecting Firefox 3.5, some
of which I haven't reported for lack of a readily available test case - I
don't have ready access to some of the pages involved.

So, my first question is whether the Mozilla rewrite touches the code in
Gecko that lies at the heart of the problem, and if so, whether this opens
up an opportunity for a more reliable implementation that can also lead to
simplification on the Orca side. I also have suspicions that implementing so
much of the caret navigation in Orca itself is a performance problem, as
evinced by delays that occur when navigating large and complex Web pages.

> We have a chicken and the egg problem here.  Firefox has bugs that
> keep the Orca guys from being able to make any progress.  However,
> Firefox can't be properly patched without better feedback from the
> Orca guys.  This situation has existed since June, and there has been
> close to zero progress.  I've "fixed" the current bug in Firefox with
> a patch that will probably take months or longer to be incorporated in
> any major distro.  That patch is almost certainly not quite right, and
> Orca/Firefox integration clearly has more bugs to be worked out.  If
> my patch gets integrated, and the Orca devs get access to it in let's
> say six months, and then I have to redo the patch, and wait another
> six months... well, you see the problem.
>
> I propose that we branch Firefox for now, and work together on
> integrating Orca with the patched version.  This will allow us to
> iterate on Firefox patches in days rather than months.  

I think that's the only sensible approach, as long as whoever is maintaining
the relevant parts of the code on the Mozilla side is at least monitoring the
process. We don't want a patch to emerge in the end which Mozilla developers
reject outright as unacceptable. Changes to the patch are to be expected
during review, but we need to avoid a scenario in which the patch is dismissed
out-of-hand on technical grounds.

Most of the Firefox-related Orca issues I've experienced are directly in this
area, so I'm interested in participating in any effort that would lead to a
long-term solution. I'm not an Orca developer, however, at least not yet...
but that could change at some point.