W3C home > Mailing lists > Public > w3c-wai-ua@w3.org > October to December 1998

How hard is linearisation?

From: Al Gilman <asgilman@access.digex.net>
Date: Thu, 19 Nov 1998 09:11:08 -0500 (EST)
Message-Id: <199811191411.JAA24708@access2.digex.net>
To: chuckop@microsoft.com (Charles)
Cc: w3c-wai-ua@w3.org
to follow up on what Charles said:

> Table unrolling means that the browser not use the HTML 4.0
> specified means of visually rendering the table, but to use a
> simpler means.  Outside the rendering code, you can use a
> script to manipulate the objects, but in the rendering engine
> the code is much different.

In the editor application of the core code, linearisation is
supported in writing out an HTML text.  Somewhere among those
myriad programmers, somebody already has the code to do this.

Is it really that hard to anaesthetize the table rendering and
follow the linear flow and still drive text to the screen?

Raman was quite dismissive about what Lynx does.  "Lynx doesn't
do anything to linearize tables; what it produces is just a
product of what it _doesn't_ do!"  And this seems to work.

Saying this should be done by post-processing in a script is
questionable if what needs to be done can be done simply by
eliminating steps in the prior processing.

It sounds like you are saying that the DOM is just for show, for
third-party consumption; the rendering engine doesn't use it.
Otherwise it should be simple enough to down-class tables in a
shim between the DOM and the existing rendering process.

Al

> 
> -----Original Message-----
> From: Charles McCathieNevile [mailto:charlesn@srl.rmit.EDU.AU]
> Sent: Wednesday, November 18, 1998 9:54 PM
> To: Charles (Chuck) Oppermann
> Cc: WAI UA group
> Subject: RE: today's telecon
> 
> 
> Good. So here is the sort of low-hanging fruit we are looking for, right?
> I suspect (and hope) that it is a fairly trivial extension of the approach
> to test for attributes, in the same way as CSS2 requires What I am
> thinking is that this approach can be applied to the very simplistic sort
> of table navigation (cell by cell or row by row, without context) that I
> discussed earlier. It could also solve (for example) the problem of the
> PAGL group about skipping over links which are in blocks. 
> 
> The nature of the element, and such attributes as TITLE or ALT could be 
> used to provide some meaningful feedback about the structure which is 
> being walked. It does not solve the problem of where the heck am I - for 
> tables you need a special beastie to run the algorithm that is used by 
> people looking at a table to provide that information, which is (I think) 
> also important to navigating a table.
> 
> Then I get into idler speculation - is this a technique for building 
> configurable outlines, or does the process go the other way around? And 
> so forth.
> 
> Charles McCathieNevile
> 
> On Wed, 18 Nov 1998, Charles (Chuck) Oppermann wrote:
> > Extremely easy to do.  Here's a sample JavaScript to do display the name
> of
> > the HTML tag for each element:
> > 
> > for (var i = 0, i < document.all.length, i++)
> > 	alert(document.all.item(i).tagName);
> > 
> > This is taken from page 169 of the book "Dynamic HTML" by Scott Issacs.
> > 
> > Charles Oppermann 
> > Program Manager, Accessibility and Disabilities Group,
> > Microsoft Corporation
> > mailto:chuckop@microsoft.com http://microsoft.com/enable/
> > "A computer on every desk and in every home, usable by everyone!" 
> > 
> > 
> 
Received on Thursday, 19 November 1998 09:10:02 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 27 October 2009 06:48:38 GMT