Comments on Last Call WD

2.2 Element Information Items

Please remove the [prefix] property. Namespace IIs already have a prefix
property, it is not needed on EIIs. In any case the current description of
[prefix] is broken in the case where in the lexical form the element has no
prefix but still has a non-empty namespace name ( i.e. where a default
namespace declaration is in scope ).

Please clarify the relationship between an Attribute II with a local name of
xmlns and a value of the empty string appearing in the [namespace
attributes] property and the [in-scope namespaces]. ( I know how it works,
I'm not sure the description is very clear ).

While on the subject of Attributes IIs in the [namespace attributes], these
feel really odd to me. Esp. given you've had to define a namespace to deal
with the xmlns 'prefix' ( and it doesn't really work for default namespace
decls given you end up with a prefix and no local name ). Please, please,
please, please, please remove the [namespace attributes] property and put
*back* the [declared namespaces] property ( which is a collection of
namespace information items ).


2.3 Attribute Information Items

Please remove the [prefix] property. ( same reason as for elements above )
The current description of [prefix] is also broken here because it refers to
the prefix of the element  name rather than the prefix of the attribute
name.

I'm also unconvinced about AIIs not having a [children] property. It seems
very weird that given the following xml document

<!DOCTYPE foo [
  <!ENTITY ent "Cruel">
]>
<foo bar='Goodbye &ent; World'>Goodbye &ent; World</foo>

at the infoset level I can tell that the character C, r, u, e and l in the
element content were part of an entity but I can't tell that about the
characters in the attribute. I'd like to see this be consistent. Either let
me see Entity Start/End Marker IIs in attributes ( which I think requires a
[children] property ) or remove them from element content ( which I don't
think is a viable solution ). Please not that this is *not* a request to
remove the [normalized value] property. I'd like to see *both* a [normalized
value] property and a [children] property. These two should be consistent
( e.g. the [children] property should *not* contain Character IIs for
preceding/trailing whitespace characters in the lexical form )


2.4 Processing Instruction Information Items

I'm not clear that having a [base uri] property is useful here... Why can't
we just look at the [base uri] of the parent ( EII or DI )?


2.6 Character Information Items

I'm not clear what the benefit of the [element content whitespace] property
is... Do we really need this?


2.16/2.17 CDATA Start/End Marker Information Items

I think these should be removed because ( as others have argued ) CDATA
sections are syntactic sugar.



Regards

Martin Gudgin
DevelopMentor

Received on Friday, 23 February 2001 04:19:41 UTC