Re: [g-a-devel] [Accessibility-atspi] Re: Overview of DBus ATSPI design

[Bill Haneman <gnome billhaneman ie>]

Hi Rob, Michael:

Regarding Rob's point about document interfaces, the Collection API has 
been drafted with some of these needs in mind (i.e. dealing with large 
pieces of content which may not already have been rendered is one of the 
design features).  It does, eventually, put the onus on the user agent 
to be able to satisfy its queries - which IMO is where it belongs - that 
is, unless one forces the "atk bridge" to collect this information 
piecemeal from the user agent.  This latter solution, which would be 
phase I in a phased approach to implementing Collection, has the 
advantage over the current model (where information must be collected by 
the assistive technology, via IPC) of at least happening in the process 
space of the user agent.  This "Phase I" approach still requires 
enumeration of a potentially very large document space.  The intent 
would be to then extend Collection into ATK (i.e. AtkCollection), which 
would allow the user agent to use its privileged knowledge of the 
document to fulfill the request more efficiently.  The Collection API 
also allows for (and encourages) bounded return sets - i.e. the caller 
should in most cases specify an upper limit on the number of objects 
which a given query returns.

Best regards,

Bill

Michael Meeks wrote:
> Hi Rob,
>
> On Mon, 2008-01-07 at 17:14 +0000, Rob Taylor wrote:
>   
> ...
>   
> > Ideas are appreciated for how to approach this. One possibility maybe a
> > separate 'interface' in the IPC mechanism for navigating within large
> > documents which may be only partially laid-out.
...