[
Lists Home |
Date Index |
Thread Index
]
Jeni Tennison wrote:
> > It would allow W3C Schema to be used in the same way as relax ng or
> > other schema languages that do not pollute the infoset: You might
> > want to validate the input for the sake of validation, but want the
> > stylesheet to work the same way as other occasions when you directly
> > transform the document without validating (because you know it is
> > already valid, for example).
>
> That wasn't how I interpreted what Evan was saying. He said that the
> schema would still supply the same Infoset as it would if the document
> were validated against an equivalent DTD -- including default
> attribute values and presumably default element values, given the
> definition that he gave. So I don't think that non-pollution was *his*
> reason.
There are two different kinds of "pollution" here. One is already allowed
with XSLT 1.0 and includes things like XInclude processing. The other adds
new *types* of information to the tree and includes PSVI augmentation; XSLT
1.0 doesn't know about such types. To use the terminology of XPDL Note[1],
the former is a "constructive process" and the latter is an "augmenting
process". XML Schemas does both. My proposal says nothing about enforcing or
restricting particular processes. However, it would provide a flag to
effectively disregard augmenting processes by way of restricting the data
model. "A schema that adds default element or attribute values, as well as
adding datatype information, is constructive as well as augmenting." Since
I'm not touching the processing model, I can't negate the effect of
constructive processes.
Evan
[1] http://www.w3.org/TR/xml-pipeline/#classification
|
