Separation of transport, envelope and documents

[Warren Buckley <Warren.Buckley@XIAM.com>]

Folks,

In Appendix C.4 (1639) MIME was choosen over XML as the packaging technology
and for good reasons.

A question I have; If a transport protocol optimised for XML documents
emerges in the future what impact will this have on the spec? (eg. Progress
Software have added layers to their message queing software to allow XML
documents to be sent through queues). 

Or more generally can an alternative packaging  technology be used without
impacting the whole spec?

The key to this flexibility would be to have the spec layered into the
following (almost stated in 7.1.1)

1. Header and Payload XML Document specification.
2. Packaging specification.
3. Package Transportation specification.

This is almost reflected in the current specification where

1. is defined from section 7.5 to section 8
2. A MIME implementation is specified from 7.1.3 -> 7.4
3. Is define by example for HTTP in the appendix.

My concern is that there are layer 2(packaging ) dependencies in layer 1
(the XML document) which will require changes to layer 1 when layer 2 is
changed.

Imagine in the future a transportation system that was XML based. As in the
protocol could only carry XML documents and not MIME. Or EbXML decided to
make XML encoding available in the future.

In line 502 (part of the Header XML document spec) reference is made to the
MIME Content-type of the Header Container. This external reference causes
some concern. The same requirement is already specified within the Header
Container refering inwards to the Header document in 7.3.3.2 (line 414).

There are a number of external references in the document that tightly bind
the Header document to the container rather than the other way around. This
means if an alternative container/packaging system were used then the Header
document would need to be changed.

The MIMEProblem error should probably also be generalised to packaging
errors.

The content-length issue was a layer 3 dependencies on layer 2.

My suggestion is the spec should clearly identify these layers and no
dependencies to outer layers should be created. References to outer layers
should be for example purposes only, so no mention of MIME in the XML
document specs and no mention HTTP in the MIME container spec.

Great work has gone into the specification to date and my comments are
intended to be constructive rather than critical. 

Keep up the good work,

Regards,
Warren.


XIAM Ltd.
12 Hume Street, Dublin 2, Ireland.
ph: +353 1 638 4850
fx:  +353  1 678 9070
email:  mailto://warren.buckley@xiam.com
www:  http://www.xiam.com