[sdo] [SDO-98]: Java Serialization should use XMLPath but this isnot reflected in the spec

[Radu Preotiuc-Pietro <radup@bea.com>]

Ulf von Mersewsky [23/Feb/07 03:20 PM] 
While working with the spec I found following passage in section 10 page
107:
[CONTAINMENT] indicates if property.containment is true.
...
When containment is false and the property's type is a DataObject, a URI
reference to the element containing the DataObject is used and an
sdo:propertyType declaration records the target type. Values in XML
documents will be of the form "#xpath" where the xpath is an XML path,
like the ones used in the ChangeSummary XML format described in Section
13.

Isn't this a reflection of the issue? It's not in the proper place of
the spec but it's there 
For 2.1.1 we could link to this sentence or move it to somewhere else.


Radu Preotiuc [23/Feb/07 07:24 PM] 
No, that's not it... I'm sure that there is a better place in the spec
for the statement that you point out, but this issue specifically is
about the Java serialization from chapter 6, not about the XML
documents.


Radu Preotiuc [25/Oct/07 05:16 AM] 
In the call on 10/18/07, workgroup members have verbally agreed to make
the change so that the path in section 6 (page 74 in the SDO 2.1 spec)
[path] to be defined as an XML path. The problem with that is that an
XML path may need to have prefixes declared in scope, so we need to
specify a way of declaring prefix-namespace mappings.

I propose we change section 6 as follows:

[0] [prefixListLength] ([prefix] [uri])* [path] [root]

where

[prefixListLength] is an int representing the number of prefix
declarations in scope, serialized using writeInt(prefixListLength)
[prefix] is a namespace prefix serialized using writeUTF(prefix)
[uri] is a namespace URI serialized using writeUTF(uri)

There are as many ([prefix] [uri]) pairs as the value of
prefixListLength and each ([prefix] [uri]) pair is to be interpreted as:
[prefix] is associated with [uri] for the purposes of resolving the XML
path [path].


Blaise Doughan [25/Oct/07 04:15 PM] 
Thinking more about this issue I am not convinced that we should be
capturing an XML XPath instead of an SDO XPath. The algorithm as it
stands is straight forward, unmarshal the graph/tree and execute the
XPath against the root DataObject and return the result.

Of course due to SDOs problems related to handling namespaces not all
objects will deserialize correctly:

      * A DataObject value of an open content property with a name that
        conflicts with a defined property
      * A DataObject value of an open content property with the same
        name as another open content property even though they have
        different namespaces 

These are issues that are being worked out in other JIRAs however, and
once resolved would negate the need for switching to an XML XPath.


Radu Preotiuc [03/Nov/07 01:32 AM] 
I disagree that resolving the other JIRAs will magically solve this
problem; at most, it will introduce the need to define prefix mappings
for SDO paths as well and this would still be an open issue. But, since
this interop format is optional in the spec, I would be fine, like I
said in the call, to defer this if we can't agree on the format.



Notice:  This email message, together with any attachments, may contain information  of  BEA Systems,  Inc.,  its subsidiaries  and  affiliated entities,  that may be confidential,  proprietary,  copyrighted  and/or legally privileged, and is intended solely for the use of the individual or entity named in this message. If you are not the intended recipient, and have received this message in error, please immediately return this by email and then delete it.