Resend: Re: KDE and AT-SPI [was: Re: Is it the time for "KSpeach"?]

[Bill Haneman <Bill Haneman Sun COM>]

On Wed, 2004-09-15 at 17:46, Olaf Jan Schmidt wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> [Bill Haneman, Mittwoch, 15. September 2004 18:04]
> > In any case, I think it makes more sense to use the ATK/at-spi
> > interfaces which Qt.next will be providing for this purpose,
including
> > AtkAction.
> >
> 
> I am not sure this will be technically possible, because the Qt-ATK
bridge 
> is server-side only. We don't have a solution for client side AT-SPI 
> support in Qt/KDE yet, mainly because the use of ORBit2 in KDE is not
an 
> option due to dependency issues.

Server-side we should use ATK of course.

> Our strategy for KDE Accessibility is to support AT-SPI as soon as 
> possible on the application side, and then to work with you on the 
> changes that are needed for KDE/Qt-based AT clients.
> 
> D-BUS will be used by both KDE and GNOME, so it would make sense to
use it 
> as a base protocol for AT-SPI. We could reuse most of the AT-SPI 
> protocol, of course. We might also change a number of things in the 
> AT-SPI protocol that are needed to make Gnopernicus stable and
quicker, 
> but I am no expert in this area. Thomas Friehoff mentioned a couple of
> points in his talk in Ludwigsburg.

DBUS isn't even used much in GNOME yet; I really think its a long way
from being capable (at this time) of supporting the needs of AT-SPI. 
That doesn't mean that it cannot or will not be extended later on.  All
the same, I have a concern, which I think is very realistic, that if you
implement all the features you need for a truly object-oriented,
network-capable object protocol, you will have more-or-less
reimplemented CORBA.  At least with CORBA we have a real, pre-existing
standard, and interoperability with other ORBs.

> Gnopernicus was written in such a way that it could use a different 
> protocol for AT communication, as Thomas pointed out in Ludwigsburg. 

The point is that at-spi itself could use another protocol for
communication - *if* a suitable one exists.  At this time, one does
not.  We would also need an idl compiler for any new protocol developed,
because the AT-SPI IDL is really the normative part of the interface. 
If we had an alternate protocol, IDL compiler, Python bindings, and
bindings for other languages/VMs that currently use CORBA to talk to
AT-SPI (for instance Java, which is how OpenOffice talks to at-spi),
then (and only then) we would have a drop-in solution.

> I 
> expect the situation to be similar for Dasher, since it is also
running 
> on Windows. How closely is GOK tied to CORBA? Does it use cspi?

GOK uses cspi.

Not all of our assistive technologies use cspi.  Orca for instance uses
the Python orbit bindinds.

- Bill

> Olaf
> 
> - -- 
> Olaf Jan Schmidt, KDE Accessibility Project
> KDEAP co-maintainer, maintainer of http://accessibility.kde.org
>