Re: Porting Dasher to Gnome accessibility framework

[Bill Haneman <bill haneman sun com>]

thanks Michael for making a detailed reply :-)

On Tue, 2003-02-04 at 13:00, Michael Meeks wrote:
> Hi Matthew,
> 
> On Wed, 2003-01-22 at 05:36, Matthew Garrett wrote:
> > I'm working on the Dasher project (http://www.inference.phy.cam.ac.uk).
> 
> 	Great :-)
> 
> > We've recently made our 3.0 release, which marks the point where we're
> > developing it fairly seriously. There's currently a GTK interface, but
> > we're looking at integrating things more usefully into Gnome. There's a
> > few questions, though:
> >
> > 1) In its current form, Dasher doesn't provide anything in the way of
> > direct interface with applications. Modifying it to enter text into the
> > selected window/textbox is pretty easy. Would it be worth it to try to
> > integrate more closely, in a similar way to GOK? The model really isn't
> > optimised for selecting menu options or similar activities, though there's
> > no real reason why this couldn't be kludged on in some fashion.

I think that Dasher in 'free text mode' would be very useful; I note
that Dasher does various kinds of bifurcation-type dynamic presentation
of selections (though since there are more than 2 choices at a time
maybe it's not "bi"-furcation ;-)  It seems to me that Dasher could
present non-text branches in the selection stream, representing things
like "menu", "grab UI", "toolbar", "back", etc... in other words, the
branching process that GOK users now undertake when they select various
of GOK's virtual keyboards could be done "Dasher-style" rather than via
virtual "keys".  Have you had a chance to run GOK in dwell mode with
GNOME-2 yet?  That might give you a good idea of how
automatic/timer-based branching in GOK works and be good food for
thought with regard to Dasher integration.

For instance if Dasher's stream during free-form text input is taken to
look sort of like (forgive the crude rendition)

--ar -- area
      --are
       -arg
	...
(etc.)
 
then after a word and whitespace has been entered, it might look like

---a--a 
 -----am
 --b
 --c
 ...
 -UI-
    --menu
    --toolbar
    ----dialog

(etc.) so that a Dasher user could either continue entering text, or
branch to a command or navigation mode.  This could allow Dasher users
access to a rich command set also (including the ability to manipulate
windows, etc.) as well as offering the opportunity to rapidly navigate
and interact with application UIs.

I am note sure it would require and fundamental architectural changes to
Dasher, perhaps only extensions and additions to implement output other
than free-form text.

> 	Well - I suppose this may involve changing the way how dasher works -
> so far I havn't had a chance to look at it in any detail unfortunately.
> However, if you can think of a use for detailed semantic information,
> such as what word position in a paragraph / sentence you are to inform
> that dictionary / matching search, perhaps that could help ? 

As Michael points out, the GNOME accessibility APIs do offer
considerably more contextual information about the current text
insertion point than you are probably used to; this could indeed be put
to good use by Dasher as well.

> Perhaps to
> enable menu / toolbar options as you say, I'm sure some interesting ways
> of traversing dialogs / widgets could be constructed - but possibly
> that's too far out.

I think creating a few simple branching navigation commands (next/prev,
next-window, etc.) could help Dasher users a good bit.  It's not clear
to me how Dasher users currently solve the problems of general UI
navigation as opposed to freeform text entry.

> > 2) The platform independent core is written in C++ - at the moment, so is
> > the platform dependent interface code. We're using gtkmm for bindings.
> > Porting with gnomemm is presumably the easiest option for us, but in a
> > hypothetical idyllic future world where Gnome ships with various
> > accessibility applications would this be an acceptable dependency?
> 
> 	Sounds ok to me; the more interesting apps that use the C++ the
> bindings the better I think - since then they're more likely to be taken
> seriously and shipped widely.

Yes, sounds right to me as well.

regards,

Bill

> > I'll be at FOSDEM in Brussels next month, so if there's anyone involved in
> > accessibilty work going it'd be good to catch up.
> 
> 	I'll be there; I look forward to meeting you [ I'm also working nearby
> in Newmarket if you need any more detailed / hands-on assistance / pint
> drinking skills ;-] Sorry for not getting back sooner, was on holiday.
> 
> 	Regards,
> 
> 		Michael.
-- 
Bill Haneman <bill haneman sun com>