Re: GDM accessibility sans AT-SPI

[Henrik Nilsen Omma <henrik ubuntu com>]

Peter Korn wrote:
> Hi Henrik,
>
> Login is a somewhat unique environment; it is reasonable to explore 
> the question of whether one can get by without AT and AT-SPI.
>
> Unfortunately, I think the answer is "it depends upon the disability 
> need", which ends up being essentially "no, you can't really do 
> without it". 
Right, it will be the case that when you look at all the use cases (some 
of which we don't yet really support like voice recognition and 
cognitive disabilities) that you will find cases where the full AT-SPI 
solution is the best one or even an absolute requirement.

The flip side of that is that for some cases the standard AT tools 
intended for the desktop will not provide the best solution. The obvious 
case that we seem to agree on is high contrast and large fonts. Btw, I'm 
not sure I've posted our recent work on that to this list: 
http://people.ubuntu.com/~henrik/gdm (done with Stéphane Marguet)

I'm actually quite agnostic about the technology that gets used to 
provide accessibility. I've thrown out this idea so that we can have a 
discussion about the use cases. I'm not in a rush to start an alternate 
development path though I am in a bit of a rush to provide workable 
solutions to our users. With the AT-SPI solution I don't have a full 
overview over what is needed to make it work properly or what the 
timeframe would be. The spec I wrote presents an alternative solution 
where there is a relatively straight-forward path for some use cases, 
though admittedly not for all.

> How does that quantity of work compare to just supporting 
> AT-SPI and ATs at login? 

That is a very important question. Perhaps the next thing we should do 
is work out what exactly is missing for the AT-SPI based solution. In 
other words, make a finite list of blockers that we can start knocking down.

Ideally I'd like to see AT-SPI supported in GDM and working smoothly, 
plus custom solutions such as theming and direct use of AccessX for 
those cases where that is enough. Then the users can choose what works 
best for them.


Henrik