Draft:  draft-ietf-sipping-media-policy-dataset-04
Reviewer: Jari Urpalainen [jari.urpalainen@nokia.com]
Review Date: 26 June 2007  (XML schema review)
Review Deadline: 11 June 2007
Status: WGLC

Summary: This draft is on the right track for publication and has nits that need to be fixed before publication. 


Comments:
---------
I've looked at the schemas, so some comments. So this i-d depends on 
profile-datasets and the base schema defined there, i.e. it is not 
possible to give comments based solely on this i-d. The biggest issues 
are with extensions. So i had to revamp the base schema in uaprof quite 
a bit (changed chemas and examples are available at 
<http://validate.openlaboratory.net/policy-dataset/>). And i don't claim 
these were _ready_ but at least the examples validate.

So in uaprof I removed <ElementGenericSetting> and 
<ElementGenericSettingContainer> and added a liberal extension point 
<PropertySetExtension>  for combine="interleave" definitions (to be used 
in additional schemas that include the base schema). The model is 
similar to one used in 
<http://tools.ietf.org/wg/simple/draft-urpalainen-simple-presence-relaxng-03.txt>. 
<anyPropertySetExtension> allows easy forwards and backwards compatible 
schemas and the including schemas can restrict this extension point 
easily if they so desire (because of e.g. deterministic and unambiguous 
schemas...). At least <ElementMaxSessionBandwidth> is still missing from 
this schema as it is referenced from policy-dataset.

In policy-dataset I added "PolicyGeneralAttributes" definition because 
it was used in several places. "AttributeMediaType" definition is also 
missing, maybe it should be in uaprof. "AttributeGenericAttributes" is 
defined by "AttributeGeneric" in uaprof. These attribute extensions are 
somewhat tricky in relaxng as there may _not_ be overlapping 
definitions, but fortunately relaxng allows a very flexible but somewhat 
verbose extensions with <except> rule. Then there were some cardinality 
bugs in containers. Examples didn't include xml declarations and 
elements were not namespace qualified. The document root element in 
uaprof is <propertySet>. "ElementQoSDSCP" did have typos in references.  
Also I'm not thrilled with the convention where in section 4.1 and in 
the relaxng schema, the same content model is being described, this 
easily leads to conflicts, one example being that <session-info> has no 
extension point in the schema (i could propose a liberal one, btw., as 
relaxng does _not_ have an annoying UPA constraint ;-)). The same 
applies to <session-policy>, section 5.1.

There might be a preference for using the compact format of relaxng, so 
one could issue rnc schema as normative and perhaps rng in an appendix 
as informative.  btw. trang is capable of converting e.g. from rng to rnc.