Re: Can we please revert the removal of viewTarget attribute?

Hi Erik,

The accessibility problem with referencing a <view> without a viewTarget is
that the view only defines a rectangle within the graphic, it doesn't
indicate which particular elements (visible within that rectangle) are
relevant.

Imagine a complex map, many adjacent and overlapping elements.  A legend
has a list of features of interest, each linked to a particular view that
best displays that feature.  The views zoom in on the relevant section of
map, but that doesn't mean it will be easy for the user agent to determine
which particular elements (and therefore which alternative text) are the
focus of that view.

Although the <view id="cityView" viewTarget="city"> element itself can have
a title and description, that does not replace the potential for detailed
structured content that could be contained in the graphic itself.  If a
screen-reader follows a link to the <g id="city"> element, it will not only
make that element's title/desc available, it will also be able to read
through all the child content.  With the <view>, in contrast, once it reads
that particular element's alternative text, it would be lost in a
completely unrelated part of the document.  By having an author-defined
connection between the two, we can instruct browsers and screen readers to
follow that logical path from the view to the content.

If there is strong opposition to the viewTarget attribute, there are ARIA
attributes (aria-owns and aria-flowto) that could be used to re-direct the
reading order appropriately.  We could re-write the SVG mapping
specification accordingly and strongly recommend that authors use these
ARIA attributes.  However, that seems a little backwards when an existing
attribute in the spec serves the same purpose.

Furthermore, that would not help with #svgView target fragments.  There is
no way to add ARIA attributes, titles, or descriptions in a target
fragment.  Allowing a viewTarget parameter to the svgView allows authors to
indicate *both* a 2D rectangular target for visual users *and* an element
that is the informational target within the document structure.

All of which could be complemented by extra styling control for visual
users if the :target pseudoclass was implemented as anticipated in SVG 1.1!

I hope that helps clarify.  I would be happy to discuss more on the call
this week.

Best,
Amelia