19981105 PA minutes


  1. WC Wendy Chisholm (scribe)
  2. GV Gregg Vanderheiden
  3. JW Jason White
  4. CL Chuck Letourneau
  5. DD Daniel Dardailler
  6. CMN Charles McCathieNeville
  7. JG Jon Gunderson
  8. IJ Ian Jacobs


  1. We need to coordinate DHTML discussion with UA group. A discussion is in the works. We would like to collect issues and ideas for that discussion
  2. What is missing in this copy that we should get in?
  3. Wording for C1. It is not easy to thread the needle on that one. Looking for ideas.

Summary of action items and consensus

  1. Modify C.1 with CMN's proposed wording (with edits)
  2. Need independent reviews by other experts that have not been involved in the process. Judy has been working on a standardized letter. need someone to follow-up, send out, make this happen.
  3. We want to be consistent with the HTML 4 table algorithm with regard to when we require table markup for headers and give specific examples in techniques.
  4. Add to guidelines: group and label related links. Add to techniques: suggest span or div with a good title and class="nav" with an example.
  5. Will add "scripting" to A.10
  6. Will discuss techniques in relation to UA techniques in the Techniques document.
  7. Our next telecon will be a summit with UA to discuss overarching issues. Jon will announce to UA list.

Wording for C1

As proposed by CMN (with modifications by the editors):

Wherever possible use a W3C technology, in accordance with guidelines on its proper use. Where this is either not possible, or results in material that does not trasnform gracefully you must provide an alternative version of the content that is accessible. [Priority 1]

/* CL, JW, GV, and WC decide that CMN's proposal is good */

/* GV reads CMN wording to JG & DD for their vote */

/* JG and DD agree */

JW - basically captures.

CMN means don't use tables. list items to indent, etc.

JW but covered elsewhere and what you're referring to

CMN yes.

JW when i put out my clumsy version, suggested as a technique server-side used to substitute wherever possible. only in relation to scripts. so in techniques doc or columns.

What is missing in this copy that we should get in?

Independent disability review

look for people in community - w or w/out disabilities to go over the various guidelines to ensure haven't forgotten anything. Jutta has group on contract that could do it, JonG has students. IG too involved in process. Judy write a letter of invitation to groups to formally ask to look at guidelines. @@ need someone to follow up and kick-start the process. this happen in CG?

Simple table definition

GV don't want to deviate or conflict with HTML 4 table algorithm

JW agree, but need examples

GV e.g., multiple values in each cell. not a cell by cell data structure.

JW an e.g. of what looking for

CMN take by context

JW software following procedure in HTML

CMN poke a hole in HTML 4 table algorithm. if works, then include it as sake of consistency.

JW works, define complexity threshold where ceases to work. looking for instances where happens (table algo falls apart)

DD e.g. of HTML 4 spec is not one that doesn't work. floating headers would be a problem to the algorithm. we should point to the algorithm not preclude from e.g. of definitions.

JW and explaining circumstances when won't work by default. recent e-mail authors difficulty using HTML features when don't appear visually in test software. so need guidance about when and how to use them.

DD point to 4.0 example. or end up with long technique if provide tutorial of table complexity.

JW yes. few indications of features that cause trouble. like when first column is headers, need scope attribute.

GV checking out table in 4.0 spec of e.g. of problem discussing. /* CMN walking through where the e.g. is */ "cups of coffee consumed by each senator." why isn't that covered by the rule?

DD not sure. remember having weird headers.

GV summary: concensus do we want to be consistent with 4.0 defn?

DD wants simple as a subset, we're raising the bar for UA.

CMN if won't be implemented in while, then we need to restrict defn of simple table. communicate to UA group: need to support table algorithm. but linearize or cell by cell presentation?

JW thought that if expose markup to external module, then set aside issue then module implement algorithm.

CMN up for discussion in UA already.

GV if linearize, then support HTML 4 algorithm. try again basis in table interpretation. so if it assumes headers and not any, get gobblety-gook. therefore, use may not be sure how it supposed to be. user say try again.

JG already have that in UA.

do we want specific examples of what talking about?

scribe: I believe the consensus was that we want to be consistent with the HTML 4 table algorithm and give specific examples in techniques.

Semantic (or contextual) markup

JG e.g., links in navigation bar. use rel of link to say navbar. UA could find relationships.

CMN is rel something we want to have broadly defined.

JW appropriate keywords?

GV whether or not when have toolbar that considered a group (to present or skip over or what not) that we could wrap them and give them a name.

JW division with class.

JG if someone has site nav links then group somehow

CMN paragraph with navbar class. have links within. suggest div.

GV B section? technique under new guideline?

IJ attempt to say it?

JG you already have "group controls into semantic units (B.2)"

GV so a technique

CMN "4. group and label related links."

GV e.g. - using DIV and class=nav

CMN title on a div?

JW yes. core attribute.

GV e.g. to group links in a navbar" and not specify here in guidelines how do, push to techniques. don't want to set in stone title or class. know people want to be able to jump over them. but then we need to say something to UA.

CMN so then need standard way to do it.

/* other agreement */

CMN this is a thing we are looking to do...stay tuned....give us a good idea.

GV or state in UA that it ought to provide mechanism to skip over groups of controls.

CMN if makr up groups as span or div, give title, then ua should allow you to nav by header, links, etc. however, can nav by divs and spans?

JW yes. nav through structure. so div and span should be part of.

GV if class then user can use style sheet.

CMN style sheet, yes, title - then it ought to be exposed.

JW class for backwards compatible (for CSS-1 support)

CMN visually not a prob. to "disappear"

GV could make it disappear.

CMN push people to do both. class for ss,

JW basically agree with charles. guideline - people use feature and as suggested. class and title in techniques.

GV use both class and title.

CMN UA currently look for alt and title

JW since we don't know what's on the authors style sheet then we need to use class but doesn't affect nav feature. class for construct and title for name.

CMN then we need to think on the same lines as rel. then should we give a particular class to use?

GV could do for diff. kinds, for navbar

CMN and JW navigation

IJ nav (so that italian works as well)

GV so just "nav" (consensus). settle on div? or span?

CMN div is a block level, span in-line. are cases where need one or other.

/* some discussion about span vs. div */

GV recap, "group and label related links." in techniques" suggest span or div with title and class.

JW and define class="nav" somewhere. with an example.

Messages from Chetz

/* CMN reading from Chetz's e-mail about HTML 4 attributes not showing up in browsers at this time. so that some may or may not show up at this time */

GV so should say "if use longdesc" won't see it in your browser.

CMN therefore we should say which browser shows up in

GV we did discuss pointer to browsercaps or browserwatch.

JW basically if using particular development doc, then those features not there.

CMN cover in abstract?

JW wanted to look through in the call.

Coordination of DHTML discussion with UA group

GV a "summit" between UA and PA. so, what are the issues to focus on? maybe we are small enough to walk over to UA group when get to it. /* udpating on what did last call, a wrapper put around other things, each individually treated in guidelines. therefore, define and reference to each piece throughout guidelines. examples in techniques. most of the issues already covered. */

JG when someone developing a script - scripts in themselves are not inaccessible in themselves, onus on author to make sure that script follows UA guidelines (since it is an application). UA's could provide some repair strategies, such as ID explicit event handlers, such as to mouse events then simulate those events.

GV what kind of repair?

JG if not keyboard equivalents, the interface simulate it. i.e. nav to elements that only had onclick, mouseup, etc. would only work with explicit binding with event and element. event bubbling no repair strategies. omnibus event on body, so the event query, where was element pressed, on what element, then decide "do this." so no indication on particular element that it can respond to an event since it bubbles up to object higher in the DOM.

JW in relation to PF discussion of DOM, propose stragety: make sure CMN guideline is included so that absolutely plain that many UA don't support any scripting so the author must provide alt. Secondary, refer to UA guideline in how to make client-side more accessible for those UA's that do support. Question a problem of reference - can we refer to them if we are completed before UA.

WC concern of pointing to since don't know who support and follow

JG also concern of providing repair in UAs

JW but no repair

/* discuss if it is repair or not. CMN "repair scripting model" */

JG first step is to have authors explicitly link element to event.

CMN if guideline, "for script which are not explicityly asscoiate w/an element, explain what they do. give the user some idea." Chuck Opp's e.g. of table used as server-side imap. Onclick on table, if click in cell 1, go to x, cell2 go to y, etc. In violation of principle c1, since already w3c technology that does that. c1a - use simplest w3c technology.

GV let the user know what happen. how do?

CMN similar to good link phrases.

GV in noscript?

CMN no, in rendered content of the document.

GV i can click on any word in doc and look up in dictionary. how do i let the user know that they can do this for each word you run over. where does the phrase go?

CMN "to look up any word...click on"

GV if I right click.

IJ there seem to be scripts that don't involve user accessibility and those that do -

  1. verify form field, do I need to expose that (no interface).
  2. scripts that affect accessibility - those that trigger events, and executed when loaded. can provide noscript, but not for those bound by events. UA should provide when possible generate other means of triggering the event (mouse down, UA simulate via keyboard). still don't know how to tell the user what it will do, but it may not be necessary in all cases. when the author knows the script will cause inaccessible pg, like event bubbling, then required to provide alt. info. in the word example, create anothr form where type in the word, hit enter it looks it up. it is device dependent mechanism.

JW no standard script language, some UA don't support any. therefore anything that relied on script needs alt. bearing in mind need alt version, then what implications for PA

IJ spiel an attempt to say class of things where author needs to provide alt, complimentary info. if event handler bound to onselect (device independence) then UA can simulate. but event bubbling, back to author. alleviate authors burden by telling how to write good event handlers.

GV part in A10 - elements with own UI. discuss applets, need reflection of scripting?

JG that would be the UA perspective

JW add scripts in there and it seems to cover. overriding that many don't support script.

GV that's transform gracefully.... If does support script, and it is running, but can't use the page because script provides UI that is inoperable...

JW relegate rest of it to techniques (how to implement)

GV just modify by adding scripting after applets?

JW cross-reference to guideline re: non-w3c technologies?

GV we've just edited A10, says see A9, and that one refers to C1 if you fail. cascading guidelines. ;-)

JW think that covers it.

/* consensus to add scripts to A10 */

GV jon have other issues?

JG in techniques descriptions for explicit binding of elements to events?

JW going to change rapidly, so yes in techniques.

GV al's "take action at right place"

JG yes, fine.

GV parallel to client-side vs. server-side imaps. trying to associate action with location, vs. script someplace to take care of bubble event. what are we recommending?

JG main dependency - to make scripts accessible wherever possible. if UA guidelines to reduce device dependence. then that in techniques under ways to make scripting more accessible.

JW in addition to A10 reduce to event connected to element affected. put event-bubbling in guidelines?

GV A10 - in techniques say, "build whole interface in or at least do x, y, z to make it easier for UA" alt-text didn't help anything until the UAs supported it. so techniques follow UA guidelines. discuss what UA could do. so make accessible for UA's that support them. similar for UA - non-mouse access to scripts (guidelines) in techniques, what things specifically could be done. then two technique docs work this out after the actual guidelines are frozen. allows UA manufacturers to help come up with solutions.

JW is UA techniques going to be frozen?

JG similar structure. techniques provides support for guidelines that they are achievable. UA guidelines much more tightly scrutinized by UA developers. if they say good idea, then we have to change software.

GV they could get hauled into court in diff way than PA.

JW techniques part of recommendation and frozen? then what GV just raised addressed quickly.

JG not intended to be frozen.

/* more discussion on evolution of techniques docs */

/* consensus that next meeting will be a "summit" between UA and PA. JG will send msg. to UA list */

/* JG gave UA update - consensus - go through techniques to determine if should be done directly by ua or by 3rd party. */