comments on Appendix E

Hi all, as per my action item I've reviewed appendix E [1] (mainly from
the POV of other type systems) and here's what I found.

In the current spec, we always use the attributes named 'body' or
'headers' (in no namespace) for referencing element declarations,
whether XML Schema, DTD or Relax NG.

It means that our model of a message is one that has a single optional
body XML element information item and zero or more header XML element
information items. This isn't specified anywhere and it isn't clear if
there may be more kinds of things in a message. 

So my first suggestion is to specify an explicit language what the model
of message is, perhaps as a paragraph in the section on The Message
Reference Component. We also need to decide explicitly on the
extensibility of the message model, i.e. whether there are other things
in the model of a message. 

If we only accept XML element declarations (body and headers), it will
require that we devise a (possibly simple and limited) mapping to
non-XML stuff for use with HTTP and MIME (for exampe for URL parameters
and HTML form encoding). If we're happy with this, we will also require
that all type systems that might be used in WSDL declare XML elements
and we need to say so in the spec. I don't see that as much of a
problem, it is certainly possible for this to work with SOAP Encoding
and SOAP Data Model. It may be awkward if we have a nice non-XML data
model and a binding that uses it and we need to go through an XML
conversion step in order to describe this in WSDL.

If we accept XML element declarations and other stuff as well (i.e.
there are other kinds of stuff in a message than just header and body
XML element information items), we'll need an example for that in
Appendix E.

Hope it helps,

                   Jacek Kopecky

                   Senior Architect
                   Systinet Corporation


Received on Thursday, 18 September 2003 11:36:51 UTC