DOCBOOK-APPS: CSS Layout Complications

[ed nixon <ed.nixon@lynnparkplace.org>]

I've spent the better part of the day on the xHTML output & CSS learning 
curve; it feels a bit like a cliff at the moment. :)

It started with the question I posted this AM about removing the 
"Abstract" gentext, including the <p class="title"><br/></p> sequence 
that results if I customize the Abstract gentext string to null. I'm 
still wrestling with that.

The perplexity continued as I discovered that Internet Explorer (which 
is the preview display engine in my copy of Homesite) behaves 
differently from Mozilla when confronted with empty <div> and <h3> tags. 
I was trying to generate xhtml from the xhtml stylesheets using the 
'xml' output method and instant saxon 6.5.2.

I made the happy discovery, though still a little foggy of understanding 
why, that if I switched the output method from 'xml' to 'html', I still 
got transitional xhtml and Internet Explorer would settle down and 
behave more or less like Mozilla and as I was expecting it to behave 
against the CSS I was developing.

But, I'm still in the woods. It has to do with those empty divs, the 
empty h3 for the author information and the horizontal rule inserted by 
default after the title. In both Mozilla and IE, these empty elements 
are adding top and bottom padding and/or margins that have the effect of 
bulking up the vertical height of the page title, the section I would 
like to use as a banner title. I'd like to have better control over how 
high it is so that I can deal properly with the position of TOC div and 
the actual page content in the first class="section" division.

If I run Tidy against the file, of course it will strip out these empty, 
space generating elements and all is pretty much well. However, that's a 
new, manual and extra step in the workflow that I'd like to avoid if 
possible.

I've tried the clean-up and make valid parameters to little effect.

Does this ring any bells with folks on the list?
Have there been solutions developed?
Am I missing the silver bullet that will solve my problems by stripping 
the empty layers from my output?

If it's more appropriate to exchange experience off list, that's OK with 
me. One way or the other, I'd like to gain some insight and, better 
still, get closer to a solution.

Thanks.                       ...edN