Re: Action-1293 Proposal

Hey, Joanie!

On Wed, Mar 18, 2015 at 10:13 PM, Joanmarie Diggs <jdiggs@igalia.com> wrote:

> Hey Alex.
>
> On 03/18/2015 09:42 PM, Alexander Surkov wrote:
>
> >     What if there are gaps? Would you make everything explicit like this:
> >
> >        <div role="row">
> >          <span role="gridcell" aria-colindex="3">Jane</span>
> >          <span role="gridcell" aria-colindex="4">Jones</span>
> >          <span role="gridcell" aria-colindex="5">Acme, Inc.</span>
> >          <span role="gridcell" aria-colindex="9">555-1234</span>
> >        </div>
> >
> >     Or would you only indicate where the gaps are, like:
> >
> >        <div role="row">
> >          <span role="gridcell" aria-colindex="3">Jane</span>
> >          <span role="gridcell">Jones</span>
> >          <span role="gridcell">Acme, Inc.</span>
> >          <span role="gridcell" aria-colindex="9">555-1234</span>
> >        </div>
> >
> >
> > these two should be equivalent, if aria-colindex is omitted then cell
> > index equals previous cell index + 1.
>
> Yes, they should be equivalent. That isn't my question. ;) My question
> is: Should our specification provide authors with that option?
>

oh, yes. My point is the less the author touches the DOM then more
performant solution he should get.


>
> > If aria-colindex is lesser or
> > equals to previous colindex then it's ignored.
>
> Oh that reminds me, I re-read your proposed text about that. I have a
> counter proposal which I forgot to ask you about:
>
> <proposed statement>
> Authors MUST set the value for aria-colindex to an integer greater than
> or equal to 1, greater than the aria-colindex value of any previous
> elements within that same row, and less than or equal to the number of
> columns in the full table.
> </proposed statement>
>

it's good until we allow aria-colindex being omitted


>
> In other words, user agents don't ignore; authors don't give you bogus
> values in the first place. :) Thoughts?
>

either way you have to have an error section in case if the author doesn't
follow these requirements but that's probably part of UAIG.


>
> >     What if we give Jane a colleague and toss in a row span, so that the
> >     explicit values would be:
> >
> >        <div role="row">
> >          <span role="gridcell" aria-colindex="3">Jane</span>
> >          <span role="gridcell" aria-colindex="4">Jones</span>
> >          <span role="gridcell" aria-colindex="5" aria-rowspan="2">Acme,
> >     Inc.</span>
> >          <span role="gridcell" aria-colindex="9">555-1234</span>
> >        </div>
> >        <div role="row">
> >          <span role="gridcell" aria-colindex="3">John</span>
> >          <span role="gridcell" aria-colindex="4">Smith</span>
> >          <span role="gridcell" aria-colindex="9">555-1235</span>
> >        </div>
> >
> >     What would the resulting implicit/duplicate-free version look like?
> >
> >
> > This is a new scenario we didn't consider yet so far. I thought that
> > only continue set of rows or columns may be missed. So is it valid
> > scenario when each row can contain different subsets of missed rows?
>
> They are not different subsets. At least not in my mind. Both rows have
> columns 3, 4, 5, and 9. But in the first row, the cell in column 5 has a
> rowspan of 2.
>

Apparently I missed your wording about rowspan. It seems bunch of people
here (previously on other mail threads) agreed that we have to have
aria-row/colspan attributes for cell spanning and that
aria-colindex/rowindex cannot serve for that. I don't recall if you thumbed
up that though.

Having said all that.... The questions I ask are genuine questions in
> which I genuinely want your input and the input of others, and plan to
> listen to everything you all say. :) But what's admittedly running
> through my mind -- and what motivated these questions -- is this: I
> think we want to be careful about values implied as a result of author
> omission. There's too much chance for confusion and author error. And as
> you know, when authors make errors on non-HTML tables, there's an
> excellent chance we're going to wind up with completely broken
> accessible tables. CSS taught us that.
>

agree, I just have a feeling that aria-colindex may be a great source of
author errors, that's basically why I suggested to restrict it to the first
row.


>
> Mind you, I DO think we can -- and should -- make authors lives easier,
> for instance by moving the count properties to the table,


thumb up for this


> and by making
> the rowindex a property of the row (leaving it up to the user agents to
> expose it on the cell for those platforms where doing so is expected). I
> do not, however, think that colindex is a property where we should be
> doing that.
>

cell might be a good place to host colindex, we just have to be very
careful defining it.


>
> --joanie
>