Meeting minutes
BGaraventa: Issue for some years where browsers have chosen on their own which roles etc need to be traversed
https://
BGaraventa: first determine which roles should/should not be traversed
scottono: example https://
<Zakim> jamesn, you wanted to ask is this not error correction at this point?
BGaraventa: some browsers could weigh on the side of including everything
jcraig: +1 in support of jamesn language - there are opportunities for authors to get that interop to fix the issues
jcraig: allowing the engines to account for it in a flexible way allows engines to account for something which are both performant and innvative
<Zakim> jamesn, you wanted to ask if we think authors should be able to do this
BGaraventa: in principle I agree but the person who suffers is the AT user
scottono: I agree that in principle this should not be allowed
scottono: but in HTML it is allowed and this is a real thing that is happening
scottono: the rules for how name should be calculated are far more recent
<Zakim> jamesn, you wanted to react to scottono
jamesn: why can't we do that
Jamie: overriding of content only happens for cases links where they do allow name from content but name from author is also allowed
Jamie: a bunch of conversations - a bigger conflation - I don't think focusability matters much if we traverse into something
Jamie: when the root node allows name from content
Jamie: group doesn't allow Name from content - but when it is a child it will get traversed
Jamie: this focusability should get name from content but only when focusable
Jamie: this would only apply for things that allow name from content
Jamie: for roles that allow name from content there is a list of things which allow traversal
<Zakim> jcraig, you wanted to say re: "allowed" MAY or SHOULD? and to and to address the innerText comment (speccing that would not allow a better sub-set of innerText)
jcraig: perhaps should would be better
jcraig: BGaraventa sounded like you wanted to put InnerText in the spec - would be good to start with.... but wouldn't allow useragents to do a better job
jcraig: there could be better ways including using text size etc, to allow user agents to do a better job
BGaraventa: the problem is when you omit info you are determining that someone may not need to know what X is - maybe the disclaimer is important and it is smaller
jcraig: agree would be a problem. with VO you can interact - if others are suppressing that then could be a problem
Jamie: have no issue with UA using ML etc. to work something out - need to mark it such that it is a guess
Jamie: think we had this conversation at TPAC and think V. important that the user knows it is a guess
+1 to Jamie
<scottono> +1
<jcraig> VO has precedent for speaking "possibly <text>" or "probably <text>"
BGaraventa: I like Jamie idea about traversing - but not specced out clearly
<jcraig> Jamie: fine with that, but the spec should make it clear that the guess should not be the same API property as the label
<Zakim> jcraig, you wanted to throw a <table> in <button> wrench
jcraig: link is less concerning.... button leaf nodes would be suppressed. We could make this an author error - unless content is exposed in other ways
aaronlev_: good simple example is Emphasis.... we have a list of the roles in chrome
Jamie: to try to write up a proposal