Re: Infoset based rewrite of SOAP Section 4

If i interpret the intent of the infoset correctly, it is a
definitive description of the XML syntax, so that one may describe
XML without delving into the details of that syntax. The appendix
that you seek is the XML specification - XMLP should not try to
re-invent or re-document the XML syntax. Rather, it should use the
framework that has been provided - Infoset - to express how to use
it.

Otherwise, to use the most simple example, if we define an attribute
as

  <foo soap:mustUnderstand="bar">

is an implementation to assume that 
  
  <foo soap:mustUnderstand='bar'>

is illegal?

Implementors must use and understand XML syntax; otherwise, the
utility of XML is lost, and we might as well go back to defining
a purpose-built binary form.

Infoset seems to be used here to two different ends;

* to clarify the syntax of XML messages (as above)

* to leave the door open for other serializations of the Infoset in
  the future.

I'm fully behind the first. I'm not so sure about the second, but
perhaps there isn't anything that we need to immediately do to cover
it.



On Sat, Jun 30, 2001 at 12:18:58PM -0400, Rich Salz wrote:
> 
> I'm an implementor -- I've done implementations of ports 80, 88, 119,
> and 135 among others. :)  So I don't say this lightly:  rewriting the
> SOAP spec to be based on the Infoset *would be a big loss for
> implementors.*
> 
> Network protocols are not built on top of abstract "information unit"
> descriptions. They are best built by from a document that describes both
> bits on the wire -- the syntax -- and the meaning of those bits -- the
> semantics.  An infoset approach loses the first and, for many
> implementors, obscures the second in a layer of abstraction.
> 
> If there are parties that must have this information, then make it an
> appendix, possibly normative.
> 	/r$
> -- 
> Zolera Systems, Securing web services (XML, SOAP, Signatures,
> Encryption)
> http://www.zolera.com

-- 
Mark Nottingham
http://www.mnot.net/

Received on Saturday, 30 June 2001 15:07:23 UTC