[dna-issues] [JBoss JIRA] Commented: (DNA-223) XML sequencer does not properly handle namespaces of unqualified attributes

[Johnny Verhaeg (JIRA)]

    [ https://jira.jboss.org/jira/browse/DNA-223?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12429416#action_12429416 ] 

Johnny Verhaeg commented on DNA-223:
------------------------------------

The XML Sequencer is indeed written with some incorrect assumptions in mind, in particular by assuming the an XML document is always supported by a corresponding XML Schema.  Some of its behavior, while not directly supported by specification, is defensible and I'll argue should remain with a few tweaks, and some is not.  It's going to take quite a bit to explain everything we should or could be doing and why, but her goes.

First let's just look at elements in XML documents that are not backed (via either the "xsi:schemaLocation" attribute or some external means) by an XML Schema.  According to the XML Namespace spec, all non-prefixed elements belong to the default namespace, whether that namespace is specified (via xmlns='...') or not.  This is where the sequencer in its current state is definitely behaving incorrectly.  In the example above, "sourceA" and "sourceB" should both be part of the default namespace, and should not inherit the "dna" namespace of their parent ("dna:sources").  We should definitely fix this.

The attributes are a bit of a different story.  According to the spec, as Randall indicates above, non-prefixed attributes don't "automatically" get assigned any particular namespace - it's left up to interpretation "determined by the element", which pretty much means it depends upon the system that defines and/or consumes the element.  All XML entities will eventually get associated with what I'll call an "effective" namespace.  Within the scope of considering XML documents alone, I will argue the "intuitive" expectation would be that the attributes' namespaces are inherited from their containing element.  But the problem here is if document creator wants an attribute to explicitly belong to a different namespace that happens to also be the default namespace.  Looking at Randall's example again, let's assume for the moment that the "sourceA" and "sourceB" elements are both part of the DNA namespace (and prefixed accordingly).  There would be no way to force "retryLimit" (as strange as the use-case may seem) to be part of the default namespace.  The user would be forced to declare a prefix that refers to the namespace for "retryLimit" and prefix the attribute accordingly.

Now enter the world of XML Schemas.  Again, let's first look at elements.  The XSD spec introduces a scenario, which unfortunately is fairly common, that basically violates the behavior defined by the XML Namespaces spec.  If all of the following are true:
<ul>
<li>An element is defined locally within a schema as part of a complex type definition (i.e., it does not reference a globally defined element (i.e., an element definition whose parent element is a <schema> element))</li>
<li>The schema defines a "targetNamespace"</li>
<li>The "effective" form of the element is "unqualified"</li>
<li>The target namespace is <em>not</em> declared as the default namespace in the XML document (i.e., it's declared with a prefix)</li>
</ul>
then when that element appears within an XML document without a prefix, its namespace is still the target namespace.  This scenario is what the current implementation of the sequencer is incorrectly assuming, and as I stated above, should be corrected.  However, it should also still be allowed for under the given conditions.  To do so (in a post-0.2 release), we'd have to enhance the sequencer to recognize when a schema is in effect, load and parse that schema, determine whether these conditions all exist, and if so, apply the target namespace to the element(s) in question.  The good news is whether we provide this enhancement or not, at least we don't have any ambiguous scenarios.

Finally, let's look at attributes backed by a schema.  The same rules apply here as with elements, which in this case means we actually have some situations where we know what the phrase in the XML Namespaces spec, "determined by the element", means.  That's the good news.  The bad is that whereas we might choose a default behavior when we don't have a schema, such as always make non-prefixed attributes part of the default namespace, we can't necessarily do that when a schema is present and one or more of the conditions listed above <em>doesn't</em> apply - we now have an ambiguous scenario with no definitive answer.

Now let's talk a little about user expectations, using the XML Schema of Schema (SoS) as one of our examples.  The SoS is not self-defining, meaning it isn't backed by itself as its own schema (it is defined partially in terms of a DTD).  If we look, for instance, at the SoS definition for the "schema" element, we see its define specifies an attribute called "targetNamespace".  So, a typical XML Schema document (which is just a particular type of instance of an XML document), might contain the following element as its root:

<xs:schema targetNamespace="http://www.jboss.org/dna" xmlns:dna="http://www.jboss.org/dna" xmlns:xs="http://www.w3.org/2001/XMLSchema">

What would you expect the namespace of the "targetNamespace" attribute to be?  Randall and I both agree that it seems to make sense it would be part of the XML Schema namespace (referred to by the "xs" prefix in this case).  In other words, we would expect the attribute to inherit its "effective" namespace from it containing element.  In fact, in all documents this is what I presume will be the "interpretation...as determined by the element...".

So what do we do?  Keeping in mind the expectation I just discussed, I suggest that we provide configurable options for the sequencer to drive it's effective namespace resolution behavior for non-prefixed attributes, with these available options:
<ol>
<li>Always use the default namespace</li>
<li>Always inherits the namespace from the containing element</li>
<li>A "mixed-mode" option that inherits the namespace from the containing element only when no default namespace is specified, and uses the default namespace when one is specified</li>
</ol>
where option 3, the "mixed-mode" option, is the default behavior.  This last option allows for the expectations we have for the commonly encountered scenarios <em>and</em> for the "retryLimit" attribute in Randall's example to be defined as belonging explicitly to a default namespace, while still adhering to both the XML Schema and XML Namespaces specs.  These options would apply regardless of whether or not we've enhanced the sequencer to be "schema-aware" (as suggested above as a post-0.2 feature enhancement).

As a parting note, the XML Schema spec's references to the the XML Namespaces spec are all to an old version - Randall's references are all to the latest version of the XML Namespace spec.

> XML sequencer does not properly handle namespaces of unqualified attributes
> ---------------------------------------------------------------------------
>
>                 Key: DNA-223
>                 URL: https://jira.jboss.org/jira/browse/DNA-223
>             Project: DNA
>          Issue Type: Bug
>          Components: Sequencers
>    Affects Versions: 0.2
>            Reporter: Randall Hauch
>            Assignee: Randall Hauch
>             Fix For: 0.2
>
>
> I've been looking at the XML sequencer behavior, and I was a little surprised to find out that it treats unqualified attributes (i.e., those without a namespace prefix) as inheriting the namespace of the parent element.  For example, this XML document:
> <dna:system xmlns:dna="http://www.jboss.org/dna" xmlns:jcr="http://www.jcp.org/jcr/1.0">
>     <!-- Define the sources from which content is made available -->
>     <dna:sources>
>         <sourceA name="Cars" dna:classname="org.jboss.dna.connector.inmemory.InMemoryRepositorySource" retryLimit="3" />
>         <sourceB name="Aircraft" dna:classname="org.jboss.dna.connector.inmemory.InMemoryRepositorySource" />
>     </dna:sources>
> </dna:system>
> is imported so that the "retryLimit" attribute is turned into a "dna:retryLimit" property.  That property, however, should use the default namespace.
> I'm not sure whether this was intended, but I don't believe this behavior is correct.  Yes, unqualified child elements do inherit the namespace of their parent, but attributes do not.  According to the spec (http://www.w3.org/TR/xml-names/#defaulting), emphasis mine:
>     "The scope of a default namespace declaration extends from the beginning of the start-tag in which it appears to the end of the corresponding end-tag, excluding the scope of any inner default namespace declarations. In the case of an empty tag, the scope is the tag itself.
>     "A default namespace declaration applies to all unprefixed element names within its scope. Default namespace declarations do not apply directly to attribute names; the interpretation of unprefixed attributes is determined by the element on which they appear.
>     "If there is a default namespace declaration in scope, the expanded name corresponding to an unprefixed element name has the URI of the default namespace as its namespace name. If there is no default namespace declaration in scope, the namespace name has no value. The namespace name for an unprefixed attribute name always has no value. In all cases, the local name is local part (which is of course the same as the unprefixed name itself)."
> Unfortunately, none of the examples in the spec seem to show the behavior for attributes.  But here are a few links that seem to agree with my interpretation:
> http://www.twoscomplement.com/2008/03/16/xml-attribute-namespaces/
> http://annevankesteren.nl/2005/03/null-namespace

-- 
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: https://jira.jboss.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira