RE: ReSpec and how it gets used

On June 9, 2016 at 8:47:32 AM, Matt King ([email protected]) wrote:
> I don't know if this is the right thread for this comment ... if not, fee=
l free to let me know.
>
> I am an editor and I rely on the JAWS screen reader. Because Firefox has =
to be updating is
> accessibility tree as respect runs, it takes a really long time to run. I=
t is rare that
> I am able to start reading a branch in rawgit in under a minute. The ARIA=
 spec takes up to
> 2 minutes before I can read it. Then, sometimes, like today, things are v=
ery broken.

Note that Firefox is severely broken wrt accessibility. I too rely
heavily on a screen-reader and have had to stop using Firefox
altogether (and yes, I work for Mozilla... that's how terrible the
situation is).

> =C2=A0Today,
> none of the roles, states, or property sections have headings or permalin=
ks. I don't
> know if that is due to a new respec bug or a failure of respect to run co=
mpletely, or a defect
> in my spec text. Today, I know it is not a defect in my text because I ha=
ven't changed it since
> it last worked.

Problems has been with the transition to "e10s" (multi-process) in
Firefox. We are working on it - but, realistically, I don't expect
things to get better for at least 1-2 years.

> I am wondering if there is a better way for respect to work. Is there a w=
ay to make all the
> respec changes without doing it on the live DOM and then replace the enti=
re DOM or something
> like that. Content hidden with display none is left out of the AX tree, s=
o maybe the whole
> DOM could be hidden while the processing is occurring ... maybe not great=
 for everyone,
> but at least you would all have an experience that is more like mine .

That's actually a great idea. Could certainly make all changes on a
different document, then replace the document fragments.

We do exactly the above for markdown parsing. I'll investigate!

Please follow up here:
https://github.com/w3c/respec/issues/818