Re: [docbook] QUERY: URI convention in DocBook XSL Stylesheets

[Bob Foster <bob@objfac.com>]

Dave Pawson wrote:
 > At 13:01 03/10/2003 -0500, GARY Hoffman wrote:
 >> This appears to be a recurring question. But, in reviewing the
 >> comments at
 >> 
<http://www.dpawson.co.uk/xsl/sect4/uri.html>http://www.dpawson.co.uk/xsl/sect4/uri.html, 

 >> there does not appear to be a clear answer.  The question is this:
 >> What is the proper form for specifying a URI in a DocBook XSL
 >> stylesheet with respect to an xsl file located on your local hard
 >> drive? Assume the following: (1) the operating system is Windows XP
 >> Pro; (2) the xslt processor is (a) xsltproc or (b) Instant Saxon; (3)
 >> the xsl file declares UTF-8 encoding; and (4) there is no network
 >> involved.
 >
 > Its a mess.
 > http://www.dpawson.co.uk/xsl/sect4/uriIncl.html
 > Shows it even more, processor vs format.

The chart is confusing. Is it always referring to the same file at the 
same location? Are there linked files in the picture? Is "church" the 
name of a system. It's not obvious why some of these choices, which are 
clearly not URIs, are even in the chart. One entry is in it twice.

What the chart shows is that no processor handled all correct forms 
correctly, but using well-formed URIs certainly increases the chances of 
acceptance. For the two common forms, only Forethought didn't handle the 
second one correctly.

uniIncl.xml
file:///c:/sgml/uriincl/uriIncl.xml

The general outline is:

1. relativepath (relative)
2. file:///absolutepath (absolute)
3. file://authority/absolutepath (absolute)

The way I remember the RFC, the second form should work with any odd 
number of / characters after the colon, but in practice /// seems to be 
most widely accepted (Microsoft's influence?).

Processors didn't do so well with the third alternative, assuming church 
is a host name known to the network, but there might be other reasons 
for this. A simple test not in the chart is:

file://localhost/c:/sgml/uriincl/uriIncl.xml

Bob Foster