[g-a-devel]Re: questions on nsIAccessibleHyperLink

[Gilbert Fang <gilbert fang sun com>]

Bill Haneman wrote:
> On Mon, 2002-06-17 at 10:22, Gilbert Fang wrote:
> > Hi, Peter
> > Thank you very much for your suggestons.
> >
> > But for details, eg. in this case:
> >
> > I'm assuming now that the full HTML looks more like:
> >
> >   <a href="" alt="A new page"> first part <img src=""
> > alt="second image"> third part</a>
> >
> > Link: "A new page" at "first part [image: second part] third part"
> >         going to "newPage.html"
> >   ...
> >
> > (image above that each line is successively spoken or sent to Braille, and the
> > "[image: ...]" is either spoken in a different voice or indicated in Braille
> > with a special dot-pattern.
> >
> > We consider it as a HyperLink with one anchor, (mozilla did not see it 
> > as three link but three accChildren, my suggestion is to
> > see it as three but now I will change my points according to your 
> > explanation). But in ATK, we only can get ONE OBJECT
> > for one anchor. How do we expose there are really three part contents .
> >   
>
> There is no fundamental reason why the object returned by
> atk_hyperlink_get_object (hyperlink, 0) cannot be a container and thus
> have children.
Yes, I agree. I will return a nsIAccesssilbe which will have three children:
nsIAccessible(ROLE_TEXT),
  nsIAccessible(ROLE_GRAPHIC) , and nsIAccessible(ROLE_TEXT)    
> However if you choose to indicate the relationship between the text and
> images in an HTML pane by making them siblings, you have a problem since
> you then split the anchor across three different AtkHypertext objects; I
> think this is the main issue.
I do make them silblings, but not by three different nsIAccessibleHyperText
object. Actually, they are nsIAceessible object.
So I do not think it is a problem . 
> If on the other hand, you consider the entire paragraph or even HTML
> page to be a single AtkHypertext object, the problem becomes "where do I
> put the image" ?
Yes, in my view, every DOMNode -- that is every html elements can be seen
as a HyperText.
> If the image is a separate child of the AtkHypertext object, in this
> case it can also implement AtkHypertext, and its anchor can point to the
> same URI as the text's anchor.
Yes, the image itself is kind of HyperText, but it will not be an anchor.
It is just one part in a hyperlink.
>  This would mean that multiple anchors
> point to the same URI, but that is perhaps a reasonable interpretation
> of the user's experience.  In other words, as a user, I can 'click' on
> either the text *or* the image to follow the link, so it is not
> inconsistent with the user's experience to say that there is more than
> one anchor or hyperlink in this case.  You could have all of these
> hyperlinks return either a container that contained both AtkText and
> AtkImage, so that no matter which link is activated the user gets
> consistent information about the hyperlink itself, or you could have the
> links return an AtkObject that implemented both AtkImage *and* AtkText.
Yes. I see. I think a container that contained both ATKText and ATKImage
is more reasonable
> If you already have three children of the HTML pane, one for the first
> part of the text up to the image, one for the image, and one for the
> second part of the text, this is my suggestion:
>
> 	make all three objects implement AtkHypertext
yes, I do. 
> 	make them return the same AtkHyperlink object in all three cases
yes, I do it too now. 
> 	make atk_hyperlink_get_object (link) return a container with 		three
> children, AtkText, AtkImage, and AtkText.
yes. In fact , these are what I thought . Because I am not sure about ATK,
I need a expert on ATK  to review my thoughts.
Thank you very much Bill. 
> It requires that ATs like screenreaders query the children of objects
> returned by atk_hyperlink_get_object, if they exist, but this probably
> would need to happen anyway for some complex cases.
>
> comments are welcome, I hope this was helpful.
>
>
> -Bill
Thanks and Best Regards.
Gilbert