the infamous 3.1.1

(I am summarising several people's views here. If I have got something
wrong, please explain on-list)

Kathy suggested that we declare this as explicitly relevant to
browser-specific functions, and create a similar formulation for functions
specific to the rendered content in section 4 or 5. Although this is not
my preference, I don't strongly object.

Jon asked a bunch of pertinent questions, and suggested I try to clarify
my thinking about this guideline to the group. 

And Al said that providing an interface was not sufficient to make a
browser accessible, and was in any case handled by Assistive technology.
(That is the bit I feel most likely to have misconstrued, but here
goes...)

As I understand it the reason Jaws works so much better with MS Explorer
than with Netscape is that Explorer exposes an interface itself, whereas
Netscape relies solely on the interface exposed by the underlying
operating system. The fact that there is an interface exposed by the
underlying operating system is sufficient to expose most of the functions,
and therefore goes most of the way to satisfying 3.1.1.

Explorer does better, not only in exposing functions like the ability to
swap between frames, but in exposing features of the document, which are
not functions provided by the browser itself, through an interface. The
example of this which is probably most familiar is table reformatting -
although Explorer itself does not allow reformatting of tables, it does
provide that ability via an interface, in this case one which is built
into explorer rather than the underlying operating system (if I recall
correctly).

Where both currently fail miserably is in providing access to certain
functions related to the document. For example, it is possible to
associate an event with a List Item using the onMouseOver attribute.
However, there is currently no ability in most standalone browsers (as
opposed to browser/editors) to give the focus in a device independent
manner to a List Item. (In plain language this means that major browsers
can't put LI into the tabbing order). So far the problems are addressed by
the guideline that says (more or less) 'let users navigate to event
triggers'

But having got there (perhaps by using a tab-order - that is an
implementation-specific detail), there is currently no way for most
browsers to trigger the onMouseOver event, except with
the specific device it addresses. For Assistive technology, commonly used
to provide alternative control mechanisms, what is required is the means
to do these two things. It may be that the Operating system provides an
interface for the assistive technology to reproduce the motions of the
mouse, and that the User Agent then fires the event, because it thinks
there is a mouse over the point in question. In which case the problem is
solved. Or it may be that the User Agent itself provides such a capacity.

There are several parts to solving this kind of problem. One is that the
event model in HTML is deficient. That is a matter to go from here to
WAI Protocols and Formats, and from there to the HTML working group, at
the end of which the events will hopefully all be specified in ways which
are not tied to specific devices. (Enter onActivate as a real piece of
HTML). Another is that the Operating system, or the User Agent itself, can
provide some interface which allows the moouse to be 'mimicked' as
described above. This is not impossible - Jaws already uses some kind of
interface to provide a second cursor which can get to any part of the
rendered document and trigger a Jaws-specific event - reading.

What I have not addressed here is Jon's question about whether everyone
(Info kiosks, screen readers, etc) needs to be bound by this requirement.
The short answer is no, but I will give a longer answer in a seperate
email later tonight.

Cheers

Charles

--Charles McCathieNevile -  mailto:charles@w3.org
phone:(temporary) +1 (617) 258 8143  http://purl.oclc.org/net/charles

W3C Web Accessibility Initiative -  http://www.w3.org/WAI
545 Technology sq., Cambridge MA, USA

Received on Wednesday, 16 December 1998 15:50:57 UTC