Re: [g-a-devel] ATK - Signal indicating new AtkObject creation.

[Bill Haneman <gnome billhaneman ie>]

Peter Korn wrote:
> Another approach here is to have some call(s) (e.g. Collections 
> interface) that enables you to get a whole bunch of information based 
> on some criteria all at once.  Then, rather than saying "hey app., 
> please bring everything in this manages-descendants thing into memory 
> for me to play with", you're simply saying "hey app., please give me 
> this collection of data now".

Yes, I can confirm that this was a key motivator for Collections. 

One problem that arises time and again is the difference between 
semi-infinite spaces and small ones.  I agree with Michael that for 
"small" sets the client's expectation (i.e. the programmer, if not the 
user) is to be able to access them all programmatically, but of course 
this fails horribly if the semi-infinite space is ignored.

The other motivation for Collection was also efficiency-related - rather 
than doing lots of IPC and a huge bit dump (especially since in theory 
the processes need not be on the same physical host!), it attempts to 
transfer basic filtering tasks to the in-process component.  It also 
provided for a way of passing data sets with fewer calls.

I think these two issues are related.  It makes sense that be that the 
client only be allowed access to subsets of these semi-infinite spaces; 
however it's not clear that visibility should be the sole, or even 
primary, basis.  However if we want to make this predictable or 
consistent for the client, "nearness" may be too fuzzy a criterion.

Being able to do these sorts of intelligent searches on documents, for 
example, can be very powerful for clients - and arguably better for 
accessibility that requiring the client to understand the semantics of 
the existing application "search" features.  There may also be some 
statutory requirements around what information an app must expose to an 
assistive technology (which discussion I leave to Peter's expertise) - 
they have the potential to clash with efforts to prune what the APIs expose.

best regards

Bill

P.S. : Hi everyone!  I'm still alive... :-)
> Peter