Pete Brunet, Michael Stewart, Hiro (Hironobu Takagi), Itoh-san (Takashi Itoh), Janina Sajka, Sueann Nichols, Larry Weiss, Barry Feigenbaum, Neil Soiffer, Mike Squillace, Steve Lee, Aaron Levanthal, Cathy Laws
While waiting for a useful standalone xulrunner environment, using an extension to minefield
Minefield is the name for Mozilla trunk builds
Doing cross process calls to MSAA using Python comtypes library
Working on catching window events
MSVC8 should be OK to use (Visual Studio Express)
Steve Lee is also working in a xulrunner environment
Rob Gallo of Freedom Scientific said they would install it if this group would create an install for it
We would make it available on the opena11y IA2 site
Mick Curran tested a test version of the IA2 IDL with typelib support and it works.
A meeting will be held later in the week with Glen Gordon from FS and Victor Tsaran. Victor has a lot of experience with scripting JAWS, e.g. Firefox and many other apps running in a Mozilla environment.
Itoh-san: This is like SAPI which provides a low level interface to TTS plus the high level SpVoice object for VB programmers
Larry: In process capability is important. Asymmetric interfaces like UIA and ATK/AT-SPI are problematic.
Itoh-san: Provide a IDispatch higher level interface, like how MS provided for TTS. Microsoft provided SpVoice for .NET apps, e.g. C#, VB .NET. These are managed apps.
Alexander Surkov started. Waiting for OK from maintainer to start making changes.
Working on licensing on alphaWorks - working on common components in conjunction with aDesigner
Need a reasonable IA2 checklist, e.g could start with the Java checklist. Sueann is starting to work through that
Sueann: This is pretty high on the todo list
Mike is interested in Alexander's effort. Aaron will keep everyone up to date
Working on getting an implementation and framework
Current SWT is based on how MSAA works. However, for something like a canvas which is used to build custom controls, there is one HWND and the children are real objects. In the case of MSAA the children would be virtual. We don't need objID of client. We can have other IDs. IE does this. The canvas creates the illusion of nested controls. All the controls have the HWND of canvas, but each object has an identity. They will be real children. The net result is that instead of using a child ID, the child will be manufactured. The transient state is used to indicate this so an AT doesn't cache it. This needs to be documented.
Nothing new this week.
We need to come up with a plan/process to keep AT-SPI, IA2, and ARIA in sync.