See also: IRC log
saz: hopes everybody read requirements doc
... next step - vote on the call (or via email) about publishing as 1st WD
... 2 or 3 iterations
... then WD note
... questions or comments?
... all clear
saz: please vote now on IRC
<shadi> saz: yes
<ChrisR> yes to EARL requirements
<Sandor> sh: yes
saz: drop confidence or keep as placeholder?
... changes will be published in new WD (before summer break?)
<Zakim> niq, you wanted to say keep confidence. Having a confidence in test-description is not mutually exclusive
niq: confidence should be kept
saz: kept as is (high medium low)?
... or have different values? integer? float values?
niq: should be open to different values
saz: would be less interoparablility
... no definition for high, medium, low
... when to use a value for each test
cr: should be optional
saz: keep values, but property optional?
cr: leave open, allow to use more values
<JibberJim> I'm happy with open
jk: also high, medium, low are not really interoperable
saz: people use them, but the meaning could be different
<niq> High: this evaluation believes the assertion. Medium: we are uncertain. Low: I'm just putting this in for completeness, but you can probably ignore it
jk: imergo only uses error (100% sure) or warning(<100% sure)
saz: use cannotTell?
<Zakim> niq, you wanted to disagree with jk. confidences are properties of warning
nk: need confidence for which warning you need to worry about
<niq> no, that would be low where there's overlap
jk: cannotTell vs fail/pass with medium confidence
saz: confidence is not about warning; developers should be more strikt
<Zakim> niq, you wanted to say fail with low confidence =~ pass with low confidence =~ can't tell
nk: 'pass' is what we believe, but we are not
sure (-> low confidence)
... sometimes it's in the middle
saz: sub-classes of cannotTell?
... cmn will read minutes and comment
nk: it's not about validity
... each of properties has its own purpose
... pass or fail with confidence to aggregate test
cr: proper alt attribute: confidence is relevant
saz: there is pass or fail (sure!)
... grey zone in between
... people used pass/fail with confidence together to descibe grey zone
nk: extend validity will look clumsy
<niq> "has no alt" <-- clear fail
<niq> alt="bullet" <-- less certain fail
why not cannotTell?
nk: should not be merged with validity
cr: low confidence -> needs a look by testing person
saz: we should also use IRC
and mailing list
saz: agreement not to drop confidence
... when to use high, medium, low?
saz: leave that up to developer?
<niq> Suggest guidelines, something like - High: this evaluation believes the assertion. Medium: we are uncertain. Low: I'm just putting this in for completeness, but you can probably ignore it
saz: values from developer's own namespace
cr: ack to nk
saz: not clear enough
cr: should be optional
<CarlosI> I agree
jk: cannot be optional - one tool says fail with low confidence - another tool reads report and only says 'fail'
nk: that's why we need confidence
saz: nk, could you write a Note about confidence when testing against WCAG?
nk: not sure if have enough time
saz: others can contribute to this
... publish end of June/beginning of July
<shadi> ACTION: Nick will prepare a draft proposal for the usage of high - medium - low confidence values in the context of WCAG and bring back to the group later in June for commenting [recorded in http://www.w3.org/2005/06/07-er-minutes.html#action01]
cr: September will not be good
... prefers later
jk: beginning of october is ok
saz: f2f in first half of october