Re: [g-a-devel] [Accessibility-atspi] AT-SPI and D-Bus

[Bill Haneman <Bill Haneman Sun COM>]

On Thu, 2006-08-03 at 20:38, George Kraft wrote:
> http://lists.freestandards.org/pipermail/lsb-discuss/2006-August/002721.html
> 
> Is there a Gnome migration plan to a DBUS based at-spi-registryd?

No.  In fact the whole issue of Bonobo deprecation (yes or no) seems
still a bit open.

George, I assume you mean "at-spi", not specifically
"at-spi-registryd".  

I see a few things that would need to happen before D-BUS migration made
sense; at the moment there seems little to gain from doing so, and a
number of serious difficulties.  In theory they could of course be
addressed, but the only advantage that I can see to the migration would
be increased acceptance by KDE.  That of course would be a very positive
development which I have no wish to impede, but I have doubts about the
practical value to the end-users who should IMO be the focus of our
efforts.

The whole AT-SPI framework would need a D-BUS backend with a standard
ABI and bindings generated by an IDL compiler.  I can imagine a sort of
"meta protocol" being built for D-BUS which was specifically designed to
transport IDL-based RPC; this may make more sense than compiling from
IDL to the current D-BUS API.  From there, profiling and optimization
would be required to close the apparently large performance gap between
ORBit2 and D-BUS (it seems to come as a surprise to some people that
D-BUS is a lot slower than ORBit2 for this, but the D-BUS maintainers
have explicitly said that they haven't really optimized the code yet).

Frankly I'd rather see us putting engineering resources into other more
pressing work on the existing infrastructure, which is sorely needed. 
As I see it the D-BUS work is mostly orthogonal to end-user needs.

Bill

> 
> -- 
> 
> George (gk4)
> 
> _______________________________________________
> Accessibility-atspi mailing list
> Accessibility-atspi lists freestandards org
> http://lists.freestandards.org/mailman/listinfo/accessibility-atspi