I just received the e-mail below from Steve Zilles (member of XSL-FO
WG) as a follow-up to an issue I raised on the XSL-FO editors list. I
mentioned to him in a off-line e-mail the differences in rendering
behaviour of the various XSL-FO implementations and mentioned also
that web browsers are not much better.

Obviously he discovered the browser inconsistencies long before me using 
the test cases he attached. The test file set Steve attached can be 
viewed at http://people.apache.org/~manuel/fop/vaTest-zilles-eight.htm
Try those out on different browsers - its amazing the variety of outputs 
produced. But more importantly the reference images give an idea how 
(Steve thinks) it should be done.

The reference to Eric's website is actually quite helpful in 
understanding the CSS inline model from which the XSL-FO model is

What does this mean for us? Firstly we have engaged a member of the
XSL-FO WG in a meaningful discussion. That is very useful and I am 
thankful to Steve for taking his time on this matter. Secondly we have 
more information to compare our current inline handling against. I  
think we are pretty close but more detailed investigation could still 
show some differences between our current interpretation / 
implementation and the intended behaviour.


----------  Forwarded Message  ----------

Subject: Re: aligment-adjust property
Date: Fri, 11 Nov 2005 11:35 am
From: Steve Zilles <[EMAIL PROTECTED]>
To: Manuel Mall <[EMAIL PROTECTED]>

At 02:40 AM 11/4/2005, Manuel Mall wrote:
> Not that HTML browsers are much better. Throw a slightly unusual
> combination of divs, spans and styles at IE, Firefox and Opera and
> you get different results with each of them.

The attached ZIP file has a CSS test case that I developed while
writing the XSL spec. Based on a quick check on my part, I found that
only Opera 8.0 did the right thing.

Another good information source on what CSS is "supposed to do"is the
following file put together by Eric Meyer (who was at the time on the

[Note: Opera does not correctly handle the "middle" alignment in the
last test case in Eric's note.]

I am close to having a definition of "before-edge" for
"alignment-adjust", but need to do some fine tuning to make it correct
and as simple as possible.

Steve Zilles
115 Lansberry Court,
Los Gatos, CA 95032-4710



Reply via email to