Re: Comments about the spec

[Murata Makoto <mura034@attglobal.net>]

James Clark wrote:

> > I was at first puzzled, but you are right.
> >
> > <element name="foo"><list><empty/></list></element>
> > matches <foo>  </foo>, but <attribute
> > name="foo"><list><empty/></list></element> does not match foo="  ".
> 
> I think you meant to say:

Sorry.  I was not really awake.

> <attribute name="foo"><list><empty/></list></attribute> matches foo="  ", 
> but  <attribute name="foo"><empty/></attribute> does not.

The semantic rule for <element> applies "stripSpace" to the content, but 
that for <attribute> does not.  (As we have agreed long time ago.) "stripSpace"  
removes whitespace strings from normalized strings (i.e., no adjacent strings).

The semantic rule for <list> uses "split", which returns a sequence of tokens.  
"split" eats intervening whitespace characters.  I suppose that it also eats 
trailing and leading whitespace characters (this is not completely clear from 
the spec).

1) <element name="foo"><list><empty/></list></element> and <foo>  </foo>

Two space characters between <foo> and </foo> are removed by "stripSpace".  Then, 
an empty mixed-sequence is compared against <list><empty/></list>.  By applying 
"split", we have an empty sequence of strings.  This matches <empty/> (6.2.3).

2) <attribute name="foo"><list><empty/></list></element> and foo="  "

A sequence of strings ("  ") is compared against <list><empty/></list>.  "split" 
eats whitespace characters, and returns an empty sequence of strings.  This matches 
<empty/>

3) <attribute name="foo"><empty/></element> and foo="  "

Neither "stripSpace" nor "split" is applied.  Thus, we examine "  ".  This 
does not match  <empty/>.

> On the other hand <element name="foo"><empty/></element> does match
> 
> <foo>  </foo>

4)  <element name="foo"><empty/></element> and <foo>  </foo>

"stripSpace" removes "  ".  Thus, we have an empty mixed-sequence.


Boy, we have many normalizations

	1) normalization by the XML processor (end-of-line and attribute value 
	normalization),

	2) normalized(m) which concatenate adjacent strings,

	3) stripSpace(m), and

	4) split(m)

I now think that some examples in the spec is helpful.  Let me try.


Cheers,

Makoto