17558 2012-06-20 11:34:50 +0000 WebVTT: syntax should disallow "-->" in cue data 2012-10-11 11:11:54 +0000 1 1 1 Unclassified TextTracks CG WebVTT unspecified PC All RESOLVED FIXED P2 normal --- 1 zcorpan ian cyril.concolato ian mike philipj public-texttracks silviapfeiffer1 dave.null oldest_to_newest 69287 0 zcorpan 2012-06-20 11:34:50 +0000 The string --> needs to be banned from the allowed syntax in cue data. Examples: <c.-->foo <v -->foo foo--> 69289 1 silviapfeiffer1 2012-06-20 12:58:21 +0000 (In reply to comment #0) > The string --> needs to be banned from the allowed syntax in cue data. Why? It might have some unexpected effects, but it's not invalid, IIUC. > Examples: > > <c.-->foo This parses to <span class="--">foo > <v -->foo This parses to <span title="--">foo > foo--> This parses to foo--> IIUC. 69290 2 zcorpan 2012-06-20 13:05:09 +0000 See step 43 in http://dev.w3.org/html5/webvtt/#webvtt-parser-algorithm (basically --> in cue data starts a new cue) 69291 3 silviapfeiffer1 2012-06-20 13:06:54 +0000 (In reply to comment #2) > See step 43 in http://dev.w3.org/html5/webvtt/#webvtt-parser-algorithm > (basically --> in cue data starts a new cue) Ah yeah, that's right, I remember. It also helps to "hide" comments. But that doesn't mean it needs banning from cue data, does it? 69292 4 zcorpan 2012-06-20 13:14:36 +0000 (In reply to comment #3) > Ah yeah, that's right, I remember. You sure what you remember isn't what the spec said before it was changed to make "-->" aggressively start a new cue? > It also helps to "hide" comments. No. > But that doesn't mean it needs banning from cue data, does it? It does. 71912 5 ian 2012-08-07 22:47:27 +0000 Oops, yeah, not sure why I forgot to do this when I changed the parser. 75897 6 contributor 2012-10-10 23:14:49 +0000 Checked in as WHATWG revision r7451. Check-in comment: Make --> in cue payloads non-conforming since it won't be parsed right anyway. http://html5.org/tools/web-apps-tracker?from=7450&to=7451 75948 7 cyril.concolato 2012-10-11 08:16:25 +0000 Hi, I have a problem with this fix. What if I carry XML data (including XML comments) in the payload of a VTT metadata cues? The --> (end of comment) will be present and since the cue is a metadata cue, it won't be escaped. 75953 8 zcorpan 2012-10-11 09:45:29 +0000 Cyril, please file a new bug. This was just making the allowed syntax match what the parser already does. 75964 9 cyril.concolato 2012-10-11 11:11:54 +0000 Done: https://www.w3.org/Bugs/Public/show_bug.cgi?id=19465