Willie Walker's Weblog

At today's meeting of the Open Accessibility (A11y) Group, Mark Doffman and Rob Taylor took center stage with a discussion of their investigation of moving AT-SPI to DBus. This work was generously funded by the Mozilla Foundation, and I happen to be the mentor for it.

The reason I've been so interested in this feasibility study is that there has been a continual wave of people speaking out loudly against CORBA. In my opinion, most of the arguments are mostly political or fraught with a lack of understanding, but that's a moot point. The fact is that most of the GNOME world does seem to be moving to DBus, potentially leaving the accessibility folks out in the cold and in de facto maintainership of the CORBA stack. In addition, embedded and smaller device providers have been pretty clear in their desire to only include one IPC mechanism on their devices that use GNOME, and it will be DBus.

So, if moving AT-SPI from CORBA to DBus is feasible, it seems as though we should try. Given that Rob helps maintain DBus, Rob and Mark are probably the two most qualified people to do this investigation, and I think they went about it in a reasonable way. In today's meeting, Michael Meeks and Bill Haneman were both present. Their presence was invaluable because they were both deeply involved in creation of the current CORBA-based implementation of the AT-SPI. Both Michael and Bill asked a lot of great questions and made some great points, and it was a pleasure to hear their voices on the phone.

Overall, I think the general feeling is that we probably should continue to move forward with the next phase (doing the work), but that people want to see the design proposal fleshed out in more detail before deciding for sure. The major sticking points seem to be around things that can be tucked under the covers more or less, such as managing reference counting. There's also the question of support for the Java platform, where a new bridge will need to be written. In any case, great job by Mark and Rob. I look forward to us getting one step closer towards resolving the difficult issue of CORBA vs. DBus based upon an in depth study that produced quality information from people well versed in the technology.