From The Linux Foundation
IAccessible2 SIG Conference Call Minutes 2008/03/04
- Andres Gonzalez, Mike Squillace
- Pete Brunet, Mick Curran, Jamie Teh, Janina Sajka
- Methods that return arrays
- It was accepted that the current IDL is OK for methods that return arrays. It will be documented that the two methods that return an array of pointers to interfaces will be allocated by the client and the six which return an array of longs or BSTRs will be allocated by the server, that the max* [in] parameter will be ignored, and that the array will be allocated by the server and freed by the client with CoTaskMemFree.
- Error indicators from methods
- See bug 418025 in the FF3 bugzilla.
- These error definitions were accepted and will be documented.
- Text attributes
- It was accepted that the language attribute should be included with values formatted according to IETF RFC 3066, e.g. en-US, de-DE.
- The direction attribute could either be defined by the CSS2 direction attribute with has two values, ltr and rtl, or by the XSL spec which names the attribute writing-mode and has values of rl, lr, and tb (top to bottom, i.e. text flows from top to bottom, then left to right). It turns out that writing-mode was in an earlier version of the CSS3 Text spec but was found to be a mistake, has been taken out, and will be eventually put back in some form in a CSS3 Text Layout spec. For the next meeting consider the following:
- Is a direction attribute in any form needed by ATs?
- If so, should we wait for CSS3 to finalize?
- Adding margin-top to the already existing margin-left/right/bottom was considered but during the discussion we questioned if these attributes are needed. Consider this for the next meeting.
- UIA to IA2 bridge
- If the Novell UIA to ATK bridge was usable to create a UIA to IA2 bridge this would be good for AT devs but need to know if it could be designed and implemented such that there would not be a noticeable performance problem such as what is seen in the case of the Java Access Bridge. In the case of the Java Access Bridge the design was such that there were a lot of cross process transitions so it would be important for a UIA to IA2 bridge to minimize or eliminate that.
- NVDA is mostly out of process which is OK for focus tracking or caret tracking in a flat document. However, performance becomes an issue when having to fetch thousands of nodes. This can happen in the case of a browser, e.g. the Wikipedia WWI page has 5000 nodes. In this case NVDA had to use in process technology.
- LGPL license
- MIDL output should not be considered derivative work from the IDL. The IDL is just a spec defining method and parameter names and data types. Another IDL based on IA2 would be a derivative work. IA2 is a derivative of the OpenOffice UAAPI (UNO Accessibility API). The output of the MIDL compiler could be considered a derivative of the COM spec.
- IA2 proxy DLL and the USB key fob problem
- It would be nice to be able to have an AT on a USB key fob. This would be useful for access to public computers in universities of coffee shops. The problem is that for the IA2 interfaces to be used the interfaces have to be registered so the interfaces are known to COM. This is done by creating the IA2 proxy DLL and registering it. (The DLL is also used to define the method an parameter names and data types.) The problem is that it is not possible to register a DLL without admin privileges. I entered this as issue 125.
- AccProbe status
- 0.3.0 was released this week on eclipse.org with caret tracking, some bug fixes, and validation against the IBM internal inspect program.