[orca-list] Many Ideas, Issues and Qs: Newb, Very Long

[Veli-Pekka Tätilä <vtatila mail student oulu fi>]

Hi, I'm a long time Windows PowerUser and programmer whose been somehow
drawn to LInux for a number of years but never truely gotten things
rowling. Now I'm testing Ubuntu Hardy via Wubi and finally have FInnish
speech, custom color, speaking admin tools, Braille and windowed
magnification working. There are loads of Qs and suggestions here, so
please feel free to snip massively.

In addition to programming Perl, Lua, Java and Ruby I'm interested in
doing music via hardware MIDi, audio multi-tracking and any accessible
soft synths. Further more, I'm into the GUI power tools but not the
command-line as an end-user, so Xplorer 2 like file managers, speaking
Emacs or graphical accessible Vim, input and DSp plugin media players
that are Foobar or Winamp like and graphical tools for editing Gnome
themes for even better accessibility personally are some of the kinds of
apps that I'm potentially into. Linux seems to have a long way to go
with accessible audio software that isn't about programming.

At any rate, here are a number of suggestions and questions about Orca,
questions first. Some of these are based on thoughts occurring when
writing my MSc about the keyboard usability of GUIs, sadly in Finnish.

Questions:

q1. How do I switch the output Orca is using for the speech? It uses my
M-Audio card reserved for music where as I'd like to use my RealTEk card
dedicated for speech in stead. BOth show up in Gnome and I can even
select them, but no matter what I do, the eSPeak speech just doesnt get
transfered to the right output. On a similar note, where can I find an
accessible mixer app workinng with pulse audio? I've only found a simple
volume control slider in the Gnome panels which isn't quite enough,
since my m-audio suffers from hard panning right in Linux.

q2. Can I bind a shortcut key to launch Orca without having to use the
run box for the job? I've looked and Gnome launchers appear to have no
facility to bind hotkeys to them directly, arrtgh.

q3. How do I stay in review mode for a long time without having to hold
down the Orca key? That would be handy if having to cursor through many
lines of console. I'm using the laptop layout on a desktop machine since
I recall letters better than numbres, and as I like to touch type and
very rarely use the numpad.

q4. I'm not a big fan of Gnome, finding it annoyingly Windows like and
not power user enough, so is Xubuntu accessible with Orca? I've been
reading list archives and already know KDe, that all my sighted Linux
friends use, is not.

And now the suggestions and issues marked with s and i. Many of these
come down to taste but I'd attempt to argument my points clearly.

Issues:

i1. If you hold down tab in a dialog for a long time  using key repeat,
or do a lot of things, Orca keeps babbling stuff that's not relevant for
tens of seconds. The same goes for holding down backspace in the
terminal. Sometimes, however, this is useful. So howbout an optional
mode that flusshes any content to be spoken when-ever new content comes
about. This would give me the fastest access to the newest information.

i2. Could ORca by any chance, when you use it in Ubuntu, for example,
neatly follow right changes. So that when running as root another
instance of orca would be spawned automagically behind the scenes and
killed as well when not needed any more or after a time out. The latest
Narrator works sort of like this with UAc prompts, as an example of
OT:ish matters.

i3. I find some of the hotkeys rather problematic in Orca. I use both Us
English and Finish speech plus full screen and windowed magnification.
TO make operation fast and not dependent on the panel UI, I would really
need hotkeys for toggling lens type of magnification betweene full
screen and custom, changing the level and regardless of their distance
switching between any 2 eSpeak languages. Another example comes from
Gnome. When using the arrows to determine which columns of info to read
in the file manager I cannot use them to manage the tree branches. Are
there any other keys than plus or minus for this? Optionally requiring
the Orca key for such column movement might be worth considering, as
would a mode locking down the Orca key by user request.

i4. I find some of the chosen hotkeys rather hard to recall especially
those in review mode. It might be helpful if the physical layout of keys
would be exploited to the max. I would find it far easier if the actions
for moving lines and characters, as in a grid, would be an up, down,
left, right arrow like formation, for example. Page keys and home/ end
could maybe signify larger units. But I don't know all the orca and
Gnome conventions so this is an old MS hand speaking, <grin>.

i5. While I like eSpeak I can understand ORpheus in Win32 so much
faster. Even the fastest setting of ESpeak's Us English is not quite
fast enough for me and there's a bit too much lag for very greedily
interrupting the speech in a linear scanning task. This might be my
sound card,which supports the low latency ASIO in win32, however.

i6. I do use magnification but would find it more helpful if the focus
could be highlighted by Orca. And not just the focus of Gnome but also
where the review mode is going. Furthre more, in addition to setting no
and full scroll margins for tracking, the  ability to set arbitrary
margins in percentages in between would be very nice for heavy
full-screen magnification personally. AS would a modal thing for simply
using the mouse or hotkeys to drag and reposition the magnifier to be of
the right size. Entering its values in pixels was a bit of a mental
exercise requiring knowledge of resolution rather than reso independent
percentages.

i7. Is it just me or is Gnome access a bit clumsy these days? I sort of
like the icon view but the problem is I don't wish to view the screne
magnified all the time to get the big picture, and the icon view does
not support logical navigation. I need to be aware of how the grid is
layed out on screen to succesfully navigate it, which is not very
accessible. The emblems aren't read either, and I cannot do whild card
filtering to weed out info.

Also in the list view I'm faced with info overload and get annoyed by
the closed and expanded prompts fairly quickly. Is there no way to turn
these prompts off? ONe trick I've seen in other readers is a mode which
only tells you the change of state, not reannouncing the same state all
the time. I find that very handy in announcing selections in a multiple
selection list. Most often I implicitly select single items and don't
wish to be notified of this. yet figuring out which items of a
discontiguous selection are unselected is crucial. The third thing is
only prompting the less frequent state e.g. open, unselected. NO prompt
then implies the other state.

Verbosity is something I could talk much more about actually. To me it
is core to get the earliest info first to interrupt the speech very
rapidly. Thus I would like the order name, state, type for check boxes,
which Orca does not do. that order would give me the checked state much
faster, and even the state would tell me what type of control it is,
leading to earlier interruption. Sometimes, based on context of use, the
ability to globally toggle the permutations of state, text and role
would be extremely useful. Such as first and fore most trying to figure
out the selected items in a list, to mention but a one thing. In Win32
I've gone even further. I use the prompts yes and no alone to imply
checkboxness and supress the type.

And now some of the major suggestions that have occurred to me so far:

s1. Could ORca please support PCRe compatible regular expressions as
part of the review mode? That would give very fast matching without
requiring pasting to a good editor first. I'm a huge fan of regexp
myself since they give structure in plain text. Finding the next
unquoted line, heading based on its numbering in ASCIi text, double
caps, next paragraph no matter how many new lines and a great many
things are searches I execute all the time in a good text editor in
Win32, and in the future, probably also in Linux. Having this time
saving support at the reader level, since regexp are so universal,
rather than on the app level, would simply rock.

s2. If you do use magnification occasionally, then the logical
navigation respecting the tab order is rather inefficient at times. If a
dialog clearly makes up a grid, why not provide optional physical
navigation in that grid using the review mode using the same algorithm
you do for figure out gridness of icon views in KBd navigation? For
efficiency, the ability to optionally wrap horizontally and vertically,
or type in coords relative to any corner would be great.

s3. Another way to boost efficiency could be structural navigation not
only on the Web , but also in dialogs. Quite simply sequential wrapping
navigation of controls based on control type directly, skipping to a
different control, or going to disabled controls you don't get the focus
to. I sure would like to hit orca+b and be sure I get to all the buttons
in a dialog in an order mimicking their relative tab order.

s4. And yet some more stuff about efficiency. Consider menu navigation.
If I  know the item I'm looking for but don't know which menu or which
hotkey, I cannot directly navigate to it. Finding out the hotkey is as
slow as finding the item, too. Rather it is back to linear listening
again. If the menu is alpha sorted I might be able to figure out whether
I should cursor up or down to minimize the distance, but this is still
very inefficient.

Could orca please read underlined mnemonic keys also in the brief
setting or perhaps optionally? not only that but knowing the name of
something and not being able to navigat  based on that name is a
fundamental problem. Review mode based type ahead navigation, much like
the in-line search of FireFox, could be one solution, supported at the
screen reader, or better yet, GUI lib level.

s5. These two last suggestions have to do with software development. If
there's a lot of redundancy in text, a filter skipping the difference
between the current and previous text to be read, and optionally letting
you know of that, might be handy. Paths with a common component shown in
their entirety would be a classic example. The shell is one such thing.
Ideally the filter should offer variable strictness.

s6. Since I also do Programming, I don't personally like the way code is
read very much. This should be delt within the IDE, but since ORca runs
Python, could support for reading PYthon code be improved? The thing
that occurred to me would be, for code you know to be correct, to read
the meaning of the code based on semantics you get from a parser, rather
than a mechanical readout of the raw punctuation. The more punctuation
the handier this is, Perl would be a good example.

++$foo{i} should surely be read preinc, lookup string i in hash foo.
rather than plus plus dollar foo left brace and so on.

Another programming example. which could also be generalized for screen
reading, would be the use of audio attributes for convaying nesting
level. An example me and a friend of mine came up with would be a
singing monotone speech lisp system, that would in stead of reading you
any parens at all, sing out the operands using a pitch determined by the
nesting level and mapped to a user scale. WIth practice, and for people
with a good ear, this might maybe even work. But that is *not* all
people.

Well, hope some of these ideas of mine prove useful.
I'm not saying these are the right let alone the only way to go about
these problems I've highlighted, just ideas I've gotten rather attached
to for some reason.

PS: For some reason I've never liked PYthon with speech, else I'd
probably try to script Orca already. I've been first spoiled by and
learned to like Perl, and now that supernova in Win32 got scripting,
fell in love with the minimalist, Perl-like, spoken Lua. But let's not
get into language wars here, I even subjectively realize I hate Python
for no proper reason other than change resistance, <grin>..

And now, I'm definitely going to get some sleep.

-- 
With kind regards Veli-Pekka Tätilä (vtatila mail student oulu fi)
Accessibility, game music, synthesizers and programming:
http://www.student.oulu.fi/~vtatila