[orca-list] Why I am asking about menu item position and group size

[Joanmarie Diggs <jdiggs igalia com>]

Hey all.

Thank you very (very, very) much for your fast feedback on my question
regarding menu item position and group size. At least so far, there
seems to be an overwhelming majority of people who think that the
correct behavior is to treat the group of menu items as if there were no
visual separators -- in other words, Orca should keep doing what it is
doing. Fair enough. Now let me tell you what prompted my question.

The reason Orca is doing what you all seem to believe is the correct
thing, at least in most cases, is because Orca is deliberately ignoring
two ARIA properties, posinset and setsize, for menus and their children.
These properties are not only being used in web applications, but in
Firefox and Thunderbird itself.

To be clear, Orca does respect the posinset and setsize properties for
other object types. But with menus, the conclusion was that respecting
these two ARIA properties would be confusing for the reasons you all
provided:

* You want to Know how many items are in the menu
* Arrow navigation means a menu is just one list of items to you
* It's hard to identify you've circled around to the top in huge menus

And I'll add one more reason: Our native toolkits, like Gtk+ and Clutter
and Qt, aren't implementing ARIA properties. So if Orca starts
respecting those properties, you'll get one type of count in Firefox,
Thuderbird, and web applications and a different type of count in
everything else. If things weren't confusing before, I think that would
make them confusing.

That said, if you all had replied that Orca's current behavior is wrong
and that presenting the item position and group size should be done
based on subgroups, then the fix would be to remove the hack in Orca to
ignore posinset and setsize -- and then get Gtk+ and Clutter and Qt to
implement something similar. (Which, thanks to our having small
developer communities in which we all know each other, wouldn't be too
hard to accomplish.)

However, that isn't what you replied. From your reply, my conclusion is
that I shouldn't remove that hack. But leaving that hack in place puts
me in an interesting position: A couple of years ago, if there was some
property or behavior like posinset and setsize that struck me as wrong,
I had no problem with failing to respect it -- like what is being done
in Orca now with menus. But Igalia (my company) became a W3C member at
the end of 2013. And I became an ARIA spec co-editor in 2014. So it
would be, ahem, awkward for me to just say I disagree with these W3C
ARIA spec properties and am therefore ignoring them in Orca. <smiles>
Furthermore, there is a very strong desire in the web community (web
developers, companies with web apps, etc.) to be able to have their
applications work more or less the same regardless of platform or screen
reader being used. And I do agree completely with that desire. So....

The right thing to do seems to be the following: See if I can achieve
consensus within the W3C that position and group size should be based on
the full set of menu items; not subsets thereof. I will point to the
feedback you all provided here (thanks again!!) as evidence that this
should be considered. If that consensus can be achieved, then that can
go into the ARIA spec, the user agent implementation spec, and the
authoring guidance, and I can remove the hack from Orca without breaking
things for you. Make sense?

--joanie