Re: DOCBOOK-APPS: Docbook xsl stylesheets and accessibilityrequirements?

[ed nixon <ed dot nixon at lynnparkplace dot org>]

"Somebodyelse" responds...

Adam DiCarlo wrote:
> <snip/>
> I need to learn more about XSLT -- is this really true, can someone
> other than Tobias confirm it?  I know it's true of DSSSL but I thought
> at least technically it was possible to have one tree form transform
> into another tree form in such a way that it's possible to ensure that
> given valid inputs all possible outputs would be valid.  Maybe this is
> feasible but it's just not what XSLT provides.
I've had a difficult time generating XHTML from DocBook via the xhtml 
XSLT stylesheets, but I'm not prepared to say why at this point. I need 
to do some controlled testing with each level of XHTML and with a 
single, "typical", base document instance. I was thinking of taking a 
chapter, for example, from  the Definitive Guide.

My experience indicates that there may be inconsistencies among XSLT 
processors with respect to the handling of the xhtml namespace 
attribute. In addition, there may be problems when modified versions of 
templates are placed in the customization layer, for the same reason, 
i.e. empty xmlns attributes are added.

It's complicated for me because I'm not expert on the differences among 
the XHTML specs, the intended function of the xsl:output instruction, 
the behaviour of various browsers with respect to the presence or 
absence of a URI in the DOCTYPE declaration, and, of course, the 
behaviour of the validators themselves. For example, there are 
differences between the W3C validator and the CSE HTML Validator on the 
Windows desktop in terms of errors reported.

Some or all of this may be clear as air to others, but the various 
factors do seem to need some analysis, tabulation and documentation in 
order to reduce the smoke and FUD.

If anyone has any suggestions about a controlled test approach, I'd be 
grateful to hear them. Better still, if there is work already done, 
please let me know and I can include it in whatever I discover.

Regards.                      ...edN