W3C home > Mailing lists > Public > w3c-wai-ua@w3.org > July to September 2001

Raw minutes from 19 July 2001 UAWG teleconference

From: Ian B. Jacobs <ij@w3.org>
Date: Thu, 19 Jul 2001 16:20:19 -0400
Message-ID: <3B574103.B7CD29DC@w3.org>
To: w3c-wai-ua@w3.org
19 July 2001 UAWG teleconference

Agenda announcement: 
Participants: Jon Gunderson (Chair), Ian Jacobs (Scribe), Mickey
Quenzer, Gregory Rosmaita, Denis Anson, Al Gilman, Harvey
Bingham, David Poehlman, Jon Ferraiolo, Dean Jackson, Chris
Lilley, Jim Allan, Tim Lacy, Rich Schwerdtfeger

Regrets: Eric Hansen

Previous meeting: 12 July 2001

Next meeting: 26 July
  Regrets: JG

Reference document: 14 July draft:

0. SVG responses to UAWG comments and scheduling.

JF: We are planning to discuss the UAWG comments at teleconfs and
that will probably not happen for a couple of weeks. The SVG WG is
preoccupied with other documents right now. I think that Chris is
essential to the process.

CL: I'll be on vacation until 6 August. The SVG WG will be putting
out other documents while I'm gone.

MQ: I'll be gone first two weeks of August.

IJ: Is CL a critical resource for this?

CL: SVG WG should start without me, but I'd like to be in the loop.

JF: If CL doesn't need to be around, we can schedule an extra teleconf
next week.

CL: We will have a response on 10 August.

IJ: I am planning vacation from 10 - 25 August.

Action IJ: Return to Judy and Janet with re-scheduling.

1. Issue 516: User input pausing (Al Gilman and SVG people)

AG: I understand the issue to be the following: I felt the SVG WG
had objected to the checkpoint and that we should probably
sustain their objection. I'm looking for a way to provide
time-independent interaction that is not locked onto

CL: Suppose you have a visual presentation where you have a
steady stream of interactions possible, and they disappear from
the screen before you have interacted. You have to know which
content is in front of it (z-order).

AG: Shooting gallery example. Animals pop up. If you shoot them
when they're not behind the thicket, you score. This is all done
by ray tracing intersection and the animal and the palm
trees. Predicting when they are going to stop would be difficult.

JF: The objection I raised was that, if you look at SVG, you may
have an animate element on the animal that has it pop up. In the
same document, you might have animate motion on something else,
which is is not something you interact with.

IJ: You can tell which ones are sensitive to input through

JF: If you are animating a picture of a lion, the even that you
have that you respond to when you click.

IJ: This is like event bubbling model.

JF: E.g., the event handler might be external.

JF, CL: We share concern that implementation burden is high and
that in most cases won't be helpful. 

IJ: What about when "end" can be recognized?

JF: "end" can take a list.

DA: If you're doing a shooting gallery, you're looking at
reaction time and coordination. A shooting gallery is one thing.
But if you have decision points that come and go (e.g., a quick
time move), that's reasonable to slow down.

JF: That's a better example of what would be useful. Suppose that
you are navigating through a floor plan and the plan has
animations at various points. You can click on timed links to
change your view from time to time.

DA: What we're looking at is not that this doorway has just
passed behind a lamp and will be visible again, but that you're
about to fall off the edge.

CL: "Active duration" in SMIL animation module. If the animation
starts and stops, that's its active duration.

JF: Let's say that you can determine for a particular animation
that it has a finite active duration, how do you determine that
there might be user interaction?

DA: Is there an event handler on the element?

JF: Animations have onbegin, onend, and onrepeat. 

IJ: Where is "onmouse down" specified?

JF: On the graphics element?

JG: Is that a container? How can you tell the relationship
between the two?

JF: If an animation element has a graphical element as a parent,
and that graphical element doesn't have an href attribute, then
the animation applies to the graphic. When we export animations,
we don't use this. We put all animations outside of the things
that are being animating. And we use "href".

IJ: Is the href sufficient?

AG: You have to invert it.

JF: SVG will support xpath, so a single animation element would
point to multiple targets. And their may be indirections. You can
animate symbols and things within symbols. The thing that you are
animating may not be a root object. E.g., when you animate a
rectangle, you may not know that that rectangle can be clicked
on; if it's part of a pattern you couldn't click on it.

IJ: Is this a question of the format losing information (and
that's bad) or being so abstract as to lead to good design, but
hard on accessibility? 

JF: I think that there wasn't a priority in designing SVG for the
preservation of semantic information. There was more of a
priority given to the presentation of accurate visual rendering.

AG: I don't think that you want to go there...We went through a
long debate about g, title, and desc and we agreed that there
would be the mechanics in the language to write high-level
svg. So you could have text in objects in logical reading order
and use it in the graphic in a high-level way. The language does
not commit you to writing at a pixel-level.

JF: You can build things in SVG that are hard to recognize, and
you can also build SVG where you can apply semantics and add

IJ: Can you write accessible SVG where the author has done the
right thing and the UA can help control the timing?

JF: I would argue that the language doesn't enable this very

IJ: Does this mean that SVG today won't let authors write 
content that can't be made accessible in terms of timing?

AG: I am not happy with trying to say yes or no. I'm not
convinced that doing the tracing to what is sensitive is that
hard. I might come down on the "overlooked" side here. I think
that the specific method of pausing for each termination of
sensitivity is overly narrow. It's not one-to-one with "provide
time-independent interaction." There may be ways to achieve what
is called "equivalent facilitation". If you take a step backwards
from the individual windows of opportunity, and say "here's an
animation that accepts user input. what does the input do." 

AG: I want to go back to the case where you are essentially
picking from a list. The animation of this is that things scroll
by and you see things one at a time in a window. This is useful,
in fact, if you have a limited input vocabulary. But in standards
Window this is a combo box. Voice dialogs are friendly for people
with command-line audio style interface. In a combo box, you can
either spell it or pick it from a list. Having untimed access to
a list would be a much better access method than pausing at the
end of each opportunity.

JG: Does SVG support the additional semantics?

CL: It might be possible to derive the list algorithmically?

AG: You would probably have to put xforms in the SVG image.

JF: I would agree with this analysis. I'm a big fan of
integrating xforms and svg. Among the benefits is that provides
the semantic model for user animations. The animations are for
visual effects, but the xforms abstract model is there.

JG: So in SVG what can we do?

AG: You know that for each sensitive object that you have an
option to do something.

IJ: Can you simply ignore the animation stuff and, say, provide a
list of interactive elements that are spatially and temporally
independent? Can you walk the DOM and pull out a list of elements
with explicit event handlers and present them as a list?

JF: Adobe wouldn't implement this. In our minds the investment
that we would make would not provide much value. But we would
invest in a higher-level model. And we would tell authors to
provide semantics at a higher level. 

AG: Is there a mode in IE where IE will present some unknown XML
application in tree view (where rendering is unknown)?
(e.g,. indented table of contents).

GR: Yes. E.g., IE does this for SVG if you don't have a plug-in
to support SVG.

TL: Yes, that's true.

CL: XSL doesn't require a DTD. 

IJ: Can you solve this problem with a style sheet?

CL: Absolutely. 

AG: You could generate this view and make the sensitive things
take sensitive things asynchronously sensitive. Remember, this is
based on what you can recognize in the markup.

JG: Will SVG viewers support CSS style sheets in general?

JF: No SVG viewers today can take a CSS style sheet to change it
from graphics to text.

AG: All we know is what the user has to have access to. The user
mode might be implemented by the XML browser, not even the SVG

/* DF, JF leaves */

CL: It sounds like you're on to a good tack (style sheets). A
proposal would be useful. 

/* CL leaves */

JG: Sounds like our current proposal is to do something with this

 "Do this by pausing automatically at the end of each time
 interval where user input is possible, and resuming
 automatically after the user has explicitly completed input."

IJ: Sounds like one way of manipulating this is to re-propose it
as either (1) provide an alternative view where input is
time-independent or (2) make the interval infinite. And the UA
should provide as much context as possible (e.g., by pausing at
the end of the interval instead of the beginning).

/* RS joins, JA leaves. */

JG: Where are people with this direction?

GR: I'd like to see the proposal.
DA: I think it's the right direction.
AG: Something in this direction is probably appropriate.
TL: I'm still chewing on it. No big disagreement.
MQ: Sounds ok to me. 
DP: Creating an alternative sounds like a good way to 
    address the problems.
RS: Seems ok.
HB: Seems reasonable.

Action IJ: Talk to Dean Jackson and write a proposal for this.

/* AG leaves */

2. [Proposal] Checkpoint 10.4 highlight requirement and image maps

IJ: Since this is limited to the case of links, should be ok for

JG: Another case is using a table with a bunch of images
(mosaic), you should have to highlight the table borders.


 - Accept proposal with any necessary editorial clarifications.

3. [Proposal] Since content focus and user interface focus are required, 
make characteristics normative

HB: How much state needs to be maintained? Is it a static order
or dynamic order?

IJ: The requirement is to restore the last content focus.


 - Accept proposal with any necessary editorial clarifications.
 - Change "four" to "three" in 9.4 note:
      "Note: For example, when the user uses the "back"
      functionality, restore the *three* state variables.

4. [Clarification] Checkpoint 6.5 (alert of changes to content) does not 
apply to style changes


 - Accept proposal: alert doesn't apply to graphical rendering
 - Editorial: "alert of rendering changes" instead of 
              "alert to rendering changes".

5. [Proposal] Changes to checkpoints 4.1, 4.2, 4.3, 10.2, and 10.4

TL: The system has a color palette. You can apply to any object,

GR: You are limited by which palette you choose (e.g., high
color, true color).

TL: If you specify something that has a higher depth than what
your screen is displaying, it gets truncated. This happens at the
GDI level (graphical drivers).

JG: The other half is that the applications can query the system
for the color depth.

IJ: What's the conventional color selection mechanism in Windows?

TL: There are system metrics for various system objects. You can
have an object and declare it's type. Colors are set globally
through user preferences. In Windows, it's what's under the
Appearance mechanism.

IJ: Is there a conventional font selection dialog in Windows?

TL: Yes. 

JG: In the MS foundation classes, there is a standard dialog box
for font size, family, and text decoration.

GR: When it comes to size, the actual rendered size depends on


 - For font size, family, and text decorations, the cascade is:
   a) Conventional font selection mechanism.
   b) Conventional text drawing services.

 - For colors
   a) Conventional color selection mechanism.
   b) Range supported by graphical drivers

Action JG: Propose new language for checkpoints 4.1, 4.2, 4.3,
10.2, and 10.4 based on a cascade in each case that starts with
user dialog, then system limitations.

C. Continued Action Items

1. TL: Write to WMP people at Microsoft to get someone to attend

TL: I'm still working on this; shouldn't have any problem getting
people from WMP and IE to attend.

Ian Jacobs (ij@w3.org)   http://www.w3.org/People/Jacobs
Cell:                    +1 917 450-8783
Received on Thursday, 19 July 2001 16:21:57 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 27 October 2009 06:50:58 GMT