!sTbVRynrrTmyjabvgc:matrix.org

Riot accessibility

27 Members
Discussions on how to improve the accessibility of Riot (Web, Android, iOS)10 Servers

Load older messages


Timestamp Message
8 Nov 2019
12:54:18@pvagner:pvagner.tkPeter Vágnerokay, no problem then, we'll have to wait. Thanks for confirming.
11 Nov 2019
19:03:10@x:riot.ovhMichael (t3chguy)https://twitter.com/RiotChat/status/1193965661732642816?s=09 has a poll for whether tab trapping should be gotten rid of, people here will likely be interested
12 Nov 2019
06:08:34@marcozehe:matrix.orgMarco ZeheThat vote is at 108 participants right now, and 75% say "Yes, change it". ;)
07:13:52@pvagner:pvagner.tkPeter VágnerLooks great
11:04:35@marcozehe:matrix.orgMarco Zehe Michael (t3chguy): Thanks for your thoughtful reply to the comment in #11071. This is the way to do it: Explain why things are this way, what conventions they follow (besides the spec) etc. I especially liked the comment about alt-tab being something totally different to tab. 🙂
11:05:38@x:riot.ovhMichael (t3chguy)If I make any mistakes in what I say, for example with why a thing is bad for accessibility technology users exactly, please do correct me
11:05:43@x:riot.ovhMichael (t3chguy):)
11:07:16@x:riot.ovhMichael (t3chguy)by the way, how would you expect a roving tab index work with filtering? Should it get reset to the first element if the previously focused element is no longer available or if the list changes at all?
14:10:11@marcozehe:matrix.orgMarco Zehe
In reply to @x:riot.ovh
by the way, how would you expect a roving tab index work with filtering? Should it get reset to the first element if the previously focused element is no longer available or if the list changes at all?
I would expect to focus to the next or previous item that is closest to the one that went away. For example, if in a list of e-mails, say in Thunderbird or Apple Mail, if you select one with the keyboard and hit delete, it goes away, and the next or previous element, depending on the direction you were last arrowing in, would be selected.
14:25:37@matthew:matrix.orgMatthewon the tab-complete issue: i'm concerned that the proposal means that you have to hit tab twice to do a simple tab complete for an unambiguous item
14:26:07@matthew:matrix.orgMatthewwhereas previously you could start typing a name, hit tab, and immediately expand it out (especially if the ordering is smart)
14:26:19@matthew:matrix.orgMatthewwhich could be a really painful muscle memory change, and one that people will consider unnecessary.
14:27:51@matthew:matrix.orgMatthewwould be acceptable to solve it by inserting an invisible DOM item for screenreaders which says "Mat autocompleted to Matthew" or something
14:28:14@matthew:matrix.orgMatthewto inform screenreaders when the first tab results in an autocomplete (rather than revealing the autocomplete selector UI?)
16:47:44@marcozehe:matrix.orgMarco ZeheOh, wait... I was under the impression that an at sign would start the autocomplete, like typing the at sign and m. Arrows would then select, and tab would autocomplete. There is no need to do a tab first to start the autocomplete process. Same for emoji, which get started by a colon sign. In fact that's how the proof of concept instance worked when I tried it. In that scenario, if there's only one result, typing at sign, then m, then tab would still auto-complete as before. We could add that screen-reader-only text, but the change to being expanded is announced already, so the user knows an auto-complete result is there.
16:49:59@x:riot.ovhMichael (t3chguy)

There is no need to do a tab first to start the autocomplete process.

Indeed but there is a special case to allow Mar<TAB> anywhere without the @ sign (then Tab or Enter to confirm)
Matthew doesn't like having to press tab twice in this case

16:50:44@x:riot.ovhMichael (t3chguy) so in Matthew's vision no autocomplete prompt would come up in this case if there was only one result and Mar<TAB> would just complete it immediately without any aria cues
17:07:32@matthew:matrix.orgMatthewi agree an alternative could be to trigger name autocompletes with an @, although historically we've tried to keep the IRC semantics of completing any word into a nick if you hit tab on it (which to me makes sense, as the completed names will not be prefixed by an @)
22:36:45@x:riot.ovhMichael (t3chguy) Woo 73% (of 122 votes) for Sure, tab is for focus!
13 Nov 2019
05:35:28@marcozehe:matrix.orgMarco Zehe
In reply to @matthew:matrix.org
i agree an alternative could be to trigger name autocompletes with an @, although historically we've tried to keep the IRC semantics of completing any word into a nick if you hit tab on it (which to me makes sense, as the completed names will not be prefixed by an @)
OK, I agree that typing ma and then tab to complete to a full nick if there is only one, makes sense. To make this shine, and be better than those IRC clients who never announced this to screen readers anywhere AFAIK, a hidden ARIA cue would be great. Make it a polite live region so it doesn't barge into whatever the screen reader may be speaking, but discoverable enough that it would help and not get in the way.
17:04:57@matthew:matrix.orgMatthewright, in which case we're 100% aligned
17:05:06@matthew:matrix.orgMatthew Michael (t3chguy): does this work for you?
17:06:59@x:riot.ovhMichael (t3chguy)so only as a special case if there is one result? if there are >1 then Mat<TAB> <TAB> to select the first?
18:13:05@marcozehe:matrix.orgMarco ZeheInteresting idea. Don't know if I have ever seen it done this way, but there's always innovation. 🙂 I'd be fine with that.
18:42:45@x:riot.ovhMichael (t3chguy)I mean the way slack does it is similar for tab force-complete. Mat<TAB> up/down arrows then TAB/ENTER to confirm
19:51:38@marcozehe:matrix.orgMarco ZeheWith the difference being that with your solution, that extra up/down would not be necessary if wanting to accept the first result, right? Besides, if I start with an at sign, in Slack, the focus is immediately forced to the first autocomplete item that pops up. They probably do it through aria-activedescendant which gets set on the first item as soon as the popup opens, so the first item doesn't need to be selected via down arrow.
20:12:02@x:riot.ovhMichael (t3chguy)

With the difference being that with your solution, that extra up/down would not be necessary if wanting to accept the first result, right?

and the tab/enter also

Slack, the focus is immediately forced to the first autocomplete item that pops up. They probably do it through aria-activedescendant which gets set on the first item as soon as the popup opens, so the first item doesn't need to be selected via down arrow.

The A11Y poc pre-selects the first option also, Slack also has the force complete tab for users without @

14 Nov 2019
05:38:28@marcozehe:matrix.orgMarco ZeheOK. Many people are probably used to the Slack way by now, so mirroring this experience in Riot is OK I think.
16 Nov 2019
12:36:41@pvagner:pvagner.tkPeter VágnerI know accessibility with Chromium and Orca on linux is a new thing only available in Chromium snapshots however when caret browsing mode is turned on in chromium left and right arrow key to collapse / expand room list categories is not working. It is likelly a chromium issue rather than Riot, however when accessibility is turned off within chromium I can't test it so I'm mentioning it so perhaps some of you can test this sooner than me.
14:14:48@pvagner:pvagner.tkPeter Vágner Okay, so it appears to be a riot thing. See this treeview example. It's expandable nodes can be controlled in both firefox and chrome.

There are no newer messages yet.


Back to Room List