1. ATK1 or ATK2?

1.1. No API Change Required

• Should the information about relevant -visible and/or invisible- objects be exposed?

  • Conclusion: We need significantly improved documentation about STATE_SHOWING and STATE_VISIBLE.

  • See also: https://bugzilla.gnome.org/show_bug.cgi?id=648260#c1

  • Person Responsible:

• 551489: API doc for atk_text_get_text_before/at/after_offset() are not consistent

  • Person Responsible: Li Yuan

• Document AtkHypertext / AtkHyperlink AtkHyperlinkImpl (see bug 650118)

  • Person Responsible: MarioSanchez

• Document AtkSocket / AtkPlug (see also bug 649307)

  • Person Responsible: MarioSanchez

1.2. API Addition

• 639486: Table cells should know their index, coordinates, and spans (and possibly the parent table)

  • Conclusion: https://live.gnome.org/Hackfests/ATK2011/Agenda/table

  • Person Responsible: Li Yuan (for GAIL)

• 638924: Improve window-related events management (AtkWindow?)

  • Conclusion: We will implement AtkWindow

  • Person Responsible: AlejandroPiñeiro

• 638378: API change requested: text-selection-added, text-selection-removed, text-selection-updated

  • Conclusion: We will create AtkTextSelection (https://bugzilla.gnome.org/show_bug.cgi?id=649902) and use property-changed event

  • Person Responsible:

• 639466: API change requested: ":system" suffix

  • Conclusion: https://bugzilla.gnome.org/show_bug.cgi?id=639466#c1

  • Person Responsible: Li Yuan (for GAIL)

• 644508: Debate if a new signal "order-changed" could be useful

  • Conclusion: https://bugzilla.gnome.org/show_bug.cgi?id=644508#c3

  • Person Responsible: Li Yuan

• 644393: [atk 2]: Consider adding a document:content-changed signal

  • Conclusion: https://bugzilla.gnome.org/show_bug.cgi?id=644393#c1

  • Person Responsible:

• 649771: Consider adding attributes-changed and relation-changed events

  • Conclusion: https://bugzilla.gnome.org/show_bug.cgi?id=649771#c2

  • Person Responsible: Li Yuan

• 648259: API change requested: Provide a means to specify a "purpose" hint

  • Conclusion:: (from workgroup) https://bugzilla.gnome.org/show_bug.cgi?id=648259#c1

  • Person Responsible:

• Question: Are there roles we lack in ATK? (Current Roles)

  • Conclusion: (from workgroup) https://live.gnome.org/Hackfests/ATK2011/Agenda/Roles

  • Person Responsible:

• 649123: Create desktop-agnostic way to identify active/running ATs?

  • Conclusion: https://live.gnome.org/Hackfests/ATK2011/Agenda/client-registry

  • Person Responsible:

• 647351: consider allowing retrieval of only one relation for an object

  • Conclusion: We will implement the ability to get the relationship by type. Note that this also happens to be included in Mozilla's proposal for IA2.

  • Person Responsible: Li Yuan

• 648675: Examine IA2 for items which would improve the user experience if added to ATK

  • Conclusion: We will implement the items listed in the bug.

  • Person Responsible:

• 598952: Implement object attribute to expose toolkit/source

  • Conclusion: https://bugzilla.gnome.org/show_bug.cgi?id=598952#c6

  • Person Responsible:

• 649805: Would be good to provide a way to invalide the cache

  • Conclusion: https://bugzilla.gnome.org/show_bug.cgi?id=649805#c3

  • Person Responsible: Mike

• 345750: Add AtkCollection interface for custom implementors and large containers

  • Person Responsible:

• 326538: Add AtkTerminal

  • Person Responsible:

• 407539: ATK missing ATK_STATE_HAS_TOOL_TIP which at-spi has

  • Person Responsible:

1.3. API Break

• 639479: Improve current AtkTable hierarchy, then document best practice

  • Conclusion: https://live.gnome.org/Hackfests/ATK2011/Agenda/table

  • Person Responsible:

• 649577: atk_add_global_event_listener should only accept ATK events

  • Conclusion: https://bugzilla.gnome.org/show_bug.cgi?id=649577#c2

  • Note: Although it is true that it is an API change, it is more about the behaviour as the signature. Probably we can get it before atk2

  • Person Responsible: AlejandroPiñeiro

• 649575: Need to clarify focus management status

  • Conclusion: https://bugzilla.gnome.org/show_bug.cgi?id=649575#c1 and https://bugzilla.gnome.org/show_bug.cgi?id=649575#c2

  • Person Responsible:

• Clean up the AtkHypertext / AtkHyperlink / AtkHyperlinkImpl stuff (so AtkHyperlink could become an interface, some background 344284, bmo#343150) (see bug 650122)

  • Conclusion: To go ahead and propose get rid of the AtkHyperlinkImpl implementor interface and makes AtkHyperlink just an interface (instead of a GObject) that would be implemented by subclasses of AtkObject as needed (those instances returned as 'hyperlinks' by AtkObjects implementing the AtkHypertext interface)

  • Person Responsible: MarioSanchez

• 363439: Use CSS/XSLT names for text attribute mappings from enums

  • Conclusion: https://bugzilla.gnome.org/show_bug.cgi?id=363439#c3

  • Person Responsible:

• 642597: Study if it would be good to became AtkObject an interface

  • Conclusion: https://bugzilla.gnome.org/show_bug.cgi?id=642597#c2

  • Person Responsible:

• 647482: There are some overlapped methods on the ATK namespace

  • Person Responsible:

• 551680: in 2.0 fix AtkObjectFactory breakage ...

  • Person Responsible: Li Yuan

• 647488: Required to move to a private structure most of the Atk interfaces data

  • Person Responsible:

• 648616: ABI maintenance tasks

  • Person Responsible:

• 640625: ATK API changes are required to became fully introspectable

  • Person Responsible:

1.4. To Be Determined

• 640440: Fine-tuning event listeners

  • Conclusion: https://bugzilla.gnome.org/show_bug.cgi?id=640440#c3

  • Person Responsible:

• 644747: Debate if AtkObject "property-change" signal should be deprecated and use the current GObject "notify"

  • Conclusion: https://bugzilla.gnome.org/show_bug.cgi?id=644747#c3

  • Person Responsible:

• 640949: Review the event API and data sent with events

  • Conclusion: https://bugzilla.gnome.org/show_bug.cgi?id=640949#c1

  • See also: https://bugzilla.gnome.org/show_bug.cgi?id=649902

  • Person Responsible:

• 611507: AtkUtil problems with loading multiple Atk implementations.

  • Related: Come up with the proper way to expose object toolkit name and object toolkit version for ATK2

  • Conclusion: https://bugzilla.gnome.org/show_bug.cgi?id=611507#c3 <-- write me!

  • Person Responsible: AlejandroPiñeiro

• 647486 : Analyze accessible instantiation: factories, AtkGObjectAccessible, base object - accessible relation

  • Conclusion: https://bugzilla.gnome.org/show_bug.cgi?id=647486#c1 <-- write me!

  • Person Responsible:

• Simplifying the way key events are handled and passed from applications to ATs and 649559: AtkKeyEventStruct has several issues to solve

  • Conclusion: <-- write me!

  • Person Responsible:

• Standard accessible hierarchies for widgets (menus, combo boxes, trees, ...)

  • Current work: https://live.gnome.org/Hackfests/ATK2011/Agenda/widgets-best-practices

  • Person Responsible:

1.5. Postponed/No Additional Action Required

• 639470: API change requested: Provide a means to "group" events based on trigger

  • Conclusion: https://bugzilla.gnome.org/show_bug.cgi?id=639470#c2

• 649804: AtkAction would be better with add/remove methods

  • Conclusion: https://bugzilla.gnome.org/show_bug.cgi?id=649804#c1

• 649556: Need to start to check how eventual Wayland move would affect ATK/AT-SPI2

  • Conclusion: https://bugzilla.gnome.org/show_bug.cgi?id=649556#c1

• 647357: consider making atkrelationtype a simple C enum

  • Conclusion: https://bugzilla.gnome.org/show_bug.cgi?id=647357#c1

• 647352: cleanup the atkrelationset interface

  • Conclusion: https://bugzilla.gnome.org/show_bug.cgi?id=647352#c4

• 648131: Should we (fully) implement or eliminate the desktop abstraction?

  • Conclusion: https://bugzilla.gnome.org/show_bug.cgi?id=648131#c1

2. Performance

• Mission: allow to have the a11y toolkit enabled by default without performance loss
  • Fits here or on an gtk hackfest
• From here: ThreePointOne/Features/Accessibility
  • "To make headway on a good accessibility story, we need to get to the point where we can turn on toolkit accessibility by default without adverse performance effects. "

• One already discussed proposal: https://mail.gnome.org/archives/gnome-accessibility-devel/2011-February/msg00037.html

• Related to bug 649123

3. Releasing

• How many thing breaks ATK/AT-SPI2 API?
  • Release plan for ATK2?
• As it is a ATK break, it would be required to update implementors ("GAIL2", "Cally2") etc.
  • Release plan for GAIL2, Cally2

  • For gtk3 or gtk4? => we don't know gtk4 dates

  • For clutter or clutter2? => we don't know clutter2 dates

• Other modules
• Maintain support for both simultaneously?
• Issues related with update at-spi2 API and how know if implementors implements that
  • Consider add a AT-SPI2 method to expose the version
• Who will do each thing?

• Conclusions

  • It is too early to make a final schedule for ATK2
  • We have created a target milestone on bugzilla to classify bugs breaking API
  • About the modules affected:
    • For the case of gtk and clutter it seems that the best idea is targetting gtk4 and clutter2
    • In the case of firefox it is not as important
    • WebkitGTK probably would do something similar
  • Frederik already created a bug to track the "AT-SPI2 exposing the version thing"
  • As we said, too early for a final conclusion, mostly because we would start with the non-breaking API tasks