Re: How should at-spi be turned on?

[Bill Haneman <bill haneman sun com>]

On Thu, 2003-05-01 at 08:10, Michael Meeks wrote:
> Hi Peter,
> 
> On Thu, 2003-05-01 at 08:03, Peter Korn wrote:
> > However, I would like to strongly urge all development releases to have
> > accessibility turned on by default.  There are too many code paths through
> > the accessibility machinery that are getting too little exercize by the army
> > of developers out there who could be helping us find and fix bugs but aren't
> > doing so because they are using the default.
> 
> 	Until event emission is done using some limited asynchronous approach
> (as discussed with Bill), I think the performance impact of it would be
> just too dire[1]. Development builds need to be able to spot performance
> regressions too ;-)

Michael:

Asynchronous emission is actually a problem for ATs; the fact is that it
will create new, significant problems if and when we implement such an
approach, since it re-introduces some of the queue-overrun/flow-control
difficulties we encountered with async at-spi events way back when.  

I am not saying that we should reject/abandon such an approach, just
that it's not without its own difficulties.  In many ways it's nicer for
assistive technologies if the notification events block until delivery,
and may actually make for a better user experience for users who need AT
(since you don't have to discard as many events and you have some means
of flow control over the application).  But of course these are the same
characteristics that create visible performance impact when running with
AT support turned on.

I'd still like to see developers regularly testing with the
accessibility key turned on; the performance impact IMO is not so dire
as you suggest, I really only notice it in a few specific
circumstances.  However it is as you say an impediment to performance
benchmarking, you definitely want to turn it off for that unless it's AT
performance you are trying to test :-)

regards,

Bill

> 
> 	Regards,
> 
> 		Michael.
> 
> [1] - that's going to need fixing anyway but ...
> -- 
>  mmeeks gnu org  <><, Pseudo Engineer, itinerant idiot
> 
> _______________________________________________
> gnome-accessibility-list mailing list
> gnome-accessibility-list gnome org
> http://mail.gnome.org/mailman/listinfo/gnome-accessibility-list