Re: Questions about gnome-mag + orca

["Kristian Lyngstøl" <kristian bohemians org>]

On Thu, Mar 27, 2008 at 8:36 PM, Willie Walker <William Walker sun com> wrote:
>  Having said all that, we have a task for "GNOME Outreach Program:
>  Accessibility" for someone to get paid to help take the magnification
>  solution beyond where we are now.  If someone is interested in this,
>  please let us know!

I'm unsure of what the plans forward for gnome-mag is. It has been my
understanding that it is mostly in maintenance mode?

Before I go further into this, I'd like to sum up my understanding of
the magnification-related technologies involved (only listing
magnification-related tasks).

Orca:
 - Collects data on what's going on. Mouse movement. Window focus.
Cursor movement, etc.
 - Controls Gnome-mag (soon to be over DBus?)

Gnome-mag:
- Just does magnification.
- Can be composite manager, but doesn't require it.
- Uses XRender
- Conflicts with other composite managers (like Compiz).
- Has performance issues in may situations.

Compiz:
 - Must be a composite manager.
 - Provides various plugins for accessibility, including "magnifier" and "ezoom"
 - Currently requires OpenGL, plans exist for other rendering engines
(including XRender). *
 - NOT the default Gnome window manager
 - Flexible; adapting plugins or even the core for a11y is easily accomplished.
 - The hardware (driver) requirements seem to be the major blocker for
this to really take off.

Metacity:
 - Default Gnome window manager and also composite manager


With regards to Compiz, there is working going on which will make
Compiz less hardware dependant and generally better all around,
including for a11y. This is a long term effort though.

That leaves us with a gnome-mag which, in it's current state, would be
by and large useless due to performance issues. I've read Carlo's idea
about having the composite manager draw into an off screen window
belonging to an other application, but I'm not sure I'm buying it.
Might as well just tell the composite manager to magnify to the right
place to begin with.

However, making textures/window content available between applications
is also something that I know there is a lot of interest in. And some
work exist (Compatible video players can give Compiz direct access to
their textures through the Video plugin).

Depending on Compiz entirely is going to be hard to sell. I am also
unfamiliar with Metacity code. I still believe that the integration
between Orca and eZoom should be finished similar to how it was
discussed during the Accessibility Summit, but that still leaves
Gnome-mag and metacity.

I'm honestly not sure what should be done about gnome-mag. Obviously
the compositor in Metacity also complicates things, as metacity does
not currently offer any magnification, thus it basically just kills
performance for gnome-mag without giving much back.

Having gnome-mag interact with Compiz isn't really had, but then I
believe we might as well just skip Gnome-mag all together. It's
straight forward to make or adjust a Compiz plugin or two to act
exactly like Gnome-mag.

This introduces some code duplication though, which I see as the major
selling point for gnome-mag staying alive. So let's look at what
exactly Gnome-mag does, help me fill in the blanks as I have very
limited gnome-mag experience.

- Allow external control (from orca?)
- One area zoomed, an other not zoomed.
- Mark the position of the mouse

Having gnome-mag control these functions but the compositor actually
performing them seems feasable to me. This would allow gnome-mag to
retain a lot of it's logic, but the compositor to do the actual work.
This way, gnome-mag could also fall back to it's own xrender method
without a significant code duplication, and the experience should be
consistent across different composited window managers. I'm still not
quite convinced this is the way to go though.... It seems too much
like we're just striving to keep gnome-mag alive. It would, with a
composite manager, basically just filter and relay messages from Orca
or other controlling programs.

If significant work is to be done on gnome-mag, I believe we need to
rethink it's place. And what to do with Metacity. There obviously
needs to be SOME sort of interaction between a composite manager and
whatever is responsible for triggering or controlling magnification.
I'm not sure how much magnification-code would be acceptable in
metacity either, or if it's even suited for it at all.

So I guess I'm not any closer to answering my own question: What
exactly should be done with magnification?

- Kristian