Re: XML+SVG, XML+CSS, XML+XSL (was Re: Bad News on IE6 XML Support)

["Simon St.Laurent" <simonstl@simonstl.com>]

At 09:48 PM 9/8/2001 -0400, Winchel 'Todd' Vincent III wrote:
>Assuming that (1) "*still* separat[ing] content from presentation" and (2)
>"edit once, present everywhere" are the goals, and assuming further that
>what you mean above is that "content" = "(perhaps any) XML" and
>"presentation" = "SVG", then what is the difference between:
>
>XML (content) + SVG (presentation)
>
>and
>
>XML (content) + CSS (presentation)
>
>and while we're at it . . .
>
>XML (content) + XSL (presentation)

I'd suggest that these two are very similar:
XML (content) + SVG (presentation)
XML (content) + XSL-FO (presentation)

In both of those cases, having an XML-based presentation vocabulary pretty 
much demands a transformation in between.

The other choice is different:
XML (content) + CSS(presentation)

The difference involves more than the use of a non-XML syntax.  CSS is very 
different from both SVG and XSL-FO because it's not about creating a "CSS 
document".  CSS is about providing information for existing 
documents.  While you can include CSS properties in an (X)HTML style 
attribute, most of the power in the CSS approach comes from its insistence 
on decorating documents rather than being the documents.

It is possible, as Joshua Allen noted, to create "identity transforms" 
which create something else which is then decorated with CSS, but the CSS 
selectors model doesn't assume transformation in the sense of XSLT, 
creating a new document from an old one.  CSS selectors let you paint 
additional presentation information onto an XML document without modifying 
its structure.

There are costs to that approach - you can't sort a list or table, or 
perform computations - but there are also some substantial benefits.  When 
linking documents, you don't need to worry that the target of a link will 
evaporate in the transformation used to present it.  When scripting 
documents, you don't need to worry about whether you're addressing an XML 
element or some (potential multiple or zero) derivative of that element.

Decoration has less power but also avoids some of the hairier situations 
created by the use of that power.  For a lot of common document and even 
data formatting situations, CSS is both adequate and easier.  For the 
creation of SVG documents or typesetting with XSL-FO, the decoration 
approach isn't enough, and then I suggest people use higher-power tools, 
like XSLT.

I just can't bring myself to believe that using XSLT to generate (X)HTML 
from XML is necessary or useful in cases where mere decoration is all 
that's needed.

Simon St.Laurent
Associate Editor
O'Reilly & Associates, Inc.