See also: IRC log
<trackbot> Date: 08 April 2015
<janina> agenda: this
<scribe> scribenick: fesch
action-1610
<trackbot> action-1610 -- Janina Sajka to Skim css scroll snap points module level 1 to see if there are issues with content bigger than viewport http://www.w3.org/tr/css-snappoints-1/ -- due 2015-04-08 -- OPEN
<trackbot> https://www.w3.org/WAI/PF/Group/track/actions/1610
<MichaelC> action-1610 due 1 week
<trackbot> Set action-1610 Skim css scroll snap points module level 1 to see if there are issues with content bigger than viewport http://www.w3.org/tr/css-snappoints-1/ due date to 2015-04-15.
<MichaelC> The Permissions API
js: minutes from two weeks ago and last week approved
gv: recording secession
js: provides info for recording
jn: I can tell what my concerns
are - when text appears in front of gradient or images or
anything that is not a constant single color -
... what percent doesn't need to meet the color contrast ration
(ie 4.5) - for example if a line goes across the image
... Cynthia's concern - the formula for calculating the
contrast ratio from the source, not the rendering (for testing
for compliance)
gv: if text is incidental in the image it is not a problem - that is exception
... if text is over an image - then there is a problem - we prepared this in cooperation with the Light house for the blind
... by the way, it is luminosity contrast ratio - not color contrast ratio. We need to get the word color out of the name - it is just "contrast ratio" not "color contrast ratio". Color contrast may be affected by color blindness.
... We use luminosity to account for, and provide contrast for, both low vision and color blindness
... the numbers came from and are based on luminosity contrast research
... a contrast of 3:1 is usually the mainstream recommended contrast for reading text, for really big text, the mainstream contrast recommendations are less
... For low vision and color blindness to have the equivalent of 3:1 for example - you need 5:1
... Regarding "size" of text - you have to make some assumptions
- using a regular size screen, not mobile
... a 14 point font on a mobile screen will be tiny
... if the user has trouble seeing text, they have to share responsibility, that is, they need to adjust their choice of device to work for them
... someone with low vision will not be able to read small text on a small screen
... When we did the calculations for people with low vision - the numbers came in at 5:1 as the recommended value. We shifted to it 4.5:1 so you could have two sets of contrast (A can contrast with B which can contrast with C - with 5:1 that is not possible).
... with regard to text over background images - that is a problem. One solution is providing a glow around the letters. Without this, lines in the background that collide with the text could be bad since it may look like the lines are part of the text
... gradients - because it is 4.5:1 you can make that happen. A glow around letters can make them pop,
... anti
aliasing should be reapplied as it gets larger so you shouldn't get large pixelation.
<jamesn> qi gradients
gv: when you have text in an image, it is incidental and not required
... basic issue - are we making so hard for the user to comprehend the letters
... Providing a glow around the letters can sharpen the contrast over any background,
... changing the formula won't help. When you provide patterns under letters where the letters touch the patterns, it does create issues
jn: asking whether you can do it, what about the one tiny corner that doesn't meet the criteria
gv: People can use judgement to where something is material or not.
jn: when you have customers that
want to buy software, needs to meet the standard...
... some customers get excited if a few pixels don't meet the
criteria
... need some way to tell customers, that it is OK
gv: got to be some rationality ...
jn: it can take a lot of time and
effort to let folks know that you meet the spirit...
... not trival to get a UI change
gv: nothing you can do about
this...
... part of my concern that someone, a purchasing agent or
judge - not accept it, usually they go the other way
... already it talks about pixelation and antialiased edge of a
character... so do it from the center
... won't expect any reasonable evaluator would reject
something based on a random pixel
cs: one of the values of the ratio is it is algorithmic, calculable, so you can either meet the criteria or not
gv: when you do the algorithm, if you have an aliased character where does the value come from?
cs: when I do it I use the CSS (color) values
gv: you use the specified value, what about image backgrounds
cs: I tell developers to avoid image backgrounds
gv: or use glow to increase contrast
cs: our guidance is very clear, with a mathematical formula - very clear, a gradient is calculable, an alpha layer is calculable
gv: if you have a gradient, could
do a reverse gradient on the text
... we look at words, not character at a time, - low vision
users need to look at characters
... latest problem with alpha - will run into same issues -
cs: contrast - not pattern -
layering of semi transparent things, hard to know what the
color will come out..
... could update formula for transparency
gv: if the renderer can figure that out, you should be able to
fe: WCAG formula does not use alpha in contrast formula
cs: the formula in WCAG needs to be updated
<jamesn> (agrees with gv)
gv: it would be a mess to update
the formula, no reason to fix the formula
... we don't specify how you get L1
cs: I see, but it looses the simplicity of using a formula
gv: changes not needed to fomula. Just look at how to lalculate L1 L2
js: not sure what are next steps are - expect to continue... will have more things to say
<jamesn> What I really need is a way to calculate an effective colour for a background image/gradient etc.
gv: closing comments - if talking about edge case and you try to accommodate it, then you will just create another edge cases, try lightening the image to make it more readable
... over a pattern, if you don't, add a little glow around letter, can make it easier to pass
cs: thanks for coming,
gv" would be happy to come again
fe: how you get L1 value - important to know where it came from - so may need to worry about composting for transparent colors from CSS
gv: also as soon as you do this, in 6 months someone may come up with something else
jn: yes, really looking for help from background - if from an image I can do sampling -- what I want help on is how to get sampling to get a pass
gv: every environment may do things differently
cs: I want to be able to tell designers - use these colors - these pass, and do this determinsitically
gv: contrast rules do not apply to text in charts or diagrams
<jamesn> http://jdevadf.oracle.com/adf-richclient-demo/faces/components/thematicMap.jspx;jsessionid=DVeaAJbTu17imG2de_-7An1jJ3LRoXi3r5XcqFeWM1mwKzm2NsBA!1452442027 for example
gv: labels in charts and maps do need contrast