There was some discussion on where we were at. Neil forwarded the minutes from last "real" meeting on June 11.
ACTION ITEM: Janina to get a web page for the handlers group set up and Neil to post minutes there.
There was some discussion of what IA2, AT-SPI and other standards do for us. Janina thought that streams solved this. Neil felt that nothing had been specified so they really didn't. Janina clarified that she meant that a mechanism was in place and that we needed to specify how it would be used. [Ed: Subsequent email from Pete and Janina clarified that what they meant by streams was in Linux's ATK / AT-SPI and is called "Streamable Content"].
Janina asked whether we can at least know where the data start/stop is. Neil said yes for XML and not at the present for most PDF (although authors should put in a figure tag and that would do the trick).
There was a long discussion on level of granularity. Is the data meaningful in a semantic way at all of the levels that are part of the (XML) structure? Are those navigation points, etc?
Neil said maybe we should step back and ask what are we trying to do. Why do we need to worry about granularity. Why not just send data structure and the AT says "tell me how to speak it", braille, it, etc. Do we need to worry about specifying how one can navigate it?
Vladimir said that SVG is typically not very structured. So there may not be a logical way to navigate it. Typically, ViewPlus will superimpose their own structure on the SVG to give the SVG meaning.
In conclusion, in some areas there is good structure (DNA, math, music), but other areas may not have meaningful structure (eg, SVG that is just a bunch of lines).
Source: Neil Soiffer, post to Accessibilityemail@example.com