26 February 2003

1. Document Object Model Load and Save

Editors:
Johnny Stenback, Netscape
Andy Heninger, IBM (until March 2001)

Table of contents

1.1. Overview

This section defines a set of interfaces for loading and saving document objects as defined in [DOM Level 3 Core]. The functionality specified in this section (the Load and Save functionality) is sufficient to allow software developers and web script authors to load and save XML content inside conforming products. The DOM Load and Save API also allows filtering of XML content using only DOM API calls; access and manipulation of the Document is defined in [DOM Level 3 Core].

The proposal for loading is influenced by the Java APIs for XML Processing [JAXP] and by SAX2 [SAX].

1.1.1. Overview of the Interfaces

The list of interfaces involved with the Loading and Saving XML documents is:

  • DOMImplementationLS -- A new DOMImplementation interface that provides the factory methods for creating the objects required for loading and saving.
  • DOMBuilder -- A parser interface.
  • DOMInputSource -- Encapsulate information about the XML document to be loaded.
  • DOMEntityResolver -- During loading, provides a way for applications to redirect references to external entities.
  • DOMBuilderFilter -- Provide the ability to examine and optionally remove Element nodes as they are being processed during the parsing of a document.
  • DOMWriter -- An interface for writing out or serializing DOM documents.
  • DocumentLS -- Provides a client or browser style interface for loading and saving.
  • ParseErrorEvent -- ParseErrorEvent is the event that is fired if there's an error in the XML document being parsed using the methods of DocumentLS.

1.1.2. The DOMInputStream type

To ensure interoperability, the DOM Load and Save specifies the following:

Type Definition DOMInputStream

A DOMInputStream represents a reference to a byte stream source of an XML input.


IDL Definition
typedef Object DOMInputStream;

Note: Even though the DOM uses the type DOMInputStream, bindings may use different types. For example, in Java DOMInputStream is bound to the java.io.InputStream type, while in ECMAScript DOMInputStream is bound to Object.

1.1.3. The DOMOutputStream type

To ensure interoperability, the DOM Load and Save specifies the following:

Type Definition DOMOutputStream

A DOMOutputStream represents a byte stream destination for the XML output.


IDL Definition
typedef Object DOMOutputStream;

Note: Even though the DOM uses the type DOMOutputStream, bindings may use different types. For example, in Java DOMOutputStream is bound to the java.io.OutputStream type, while in ECMAScript DOMOutputStream is bound to Object.

1.1.4. The DOMReader type

To ensure interoperability, the DOM Load and Save specifies the following:

Type Definition DOMReader

A DOMReader represents a character stream for the XML input.


IDL Definition
typedef Object DOMReader;

Note: Even though the DOM uses the type DOMReader, bindings may use different types. For example, in Java DOMReader is bound to the java.io.Reader type, while in ECMAScript DOMReader is NOT bound, and therefore as no recommended meaning in ECMAScript.

1.2. Fundamental interfaces

The interface within this section is considered fundamental, and must be fully implemented by all conforming implementations of the DOM Load and Save module.

Interface DOMImplementationLS

DOMImplementationLS contains the factory methods for creating objects that implement the DOMBuilder (parser) and DOMWriter (serializer) interfaces.

The expectation is that an instance of the DOMImplementationLS interface can be obtained by using binding-specific casting methods on an instance of the DOMImplementation interface or, if the Document supports the feature "Core" version "3.0" defined in [DOM Level 3 Core], by using the method DOMImplementation.getFeature with parameter values "LS-Load" and "3.0" (respectively).


IDL Definition
interface DOMImplementationLS {

  // DOMIMplementationLSMode
  const unsigned short      MODE_SYNCHRONOUS               = 1;
  const unsigned short      MODE_ASYNCHRONOUS              = 2;

  DOMBuilder         createDOMBuilder(in unsigned short mode, 
                                      in DOMString schemaType)
                                        raises(DOMException);
  DOMWriter          createDOMWriter();
  DOMInputSource     createDOMInputSource();
};

Definition group DOMIMplementationLSMode

An integer indicating which type of mode this is.

Defined Constants
MODE_ASYNCHRONOUS
Create an asynchronous DOMBuilder.
MODE_SYNCHRONOUS
Create a synchronous DOMBuilder.
Methods
createDOMBuilder
Create a new DOMBuilder. The newly constructed parser may then be configured by means of its setFeature method, and used to parse documents by means of its parse method.
Parameters
mode of type unsigned short
The mode argument is either MODE_SYNCHRONOUS or MODE_ASYNCHRONOUS, if mode is MODE_SYNCHRONOUS then the DOMBuilder that is created will operate in synchronous mode, if it's MODE_ASYNCHRONOUS then the DOMBuilder that is created will operate in asynchronous mode.
schemaType of type DOMString
An absolute URI representing the type of the schema language used during the load of a Document using the newly created DOMBuilder. Note that no lexical checking is done on the absolute URI. In order to create a DOMBuilder for any kind of schema types (i.e. the DOMBuilder will be free to use any schema found), use the value null.

Note: For W3C XML Schema [XML Schema Part 1], applications must use the value "http://www.w3.org/2001/XMLSchema". For XML DTD [XML 1.0], applications must use the value "http://www.w3.org/TR/REC-xml". Other Schema languages are outside the scope of the W3C and therefore should recommend an absolute URI in order to use this method.

Return Value

DOMBuilder

The newly created DOMBuilder object. This DOMBuilder is either synchronous or asynchronous depending on the value of the mode argument.

Note: By default, the newly created DOMBuilder does not contain a DOMErrorHandler, i.e. the value of the errorHandler is null. However, implementations may provide a default error handler at creation time. In that case, the initial value of the errorHandler attribute on the new created DOMBuilder contains a reference to the default error handler.

Exceptions

DOMException

NOT_SUPPORTED_ERR: Raised if the requested mode or schema type is not supported.

createDOMInputSource
Create a new "empty" DOMInputSource.
Return Value

DOMInputSource

The newly created DOMInputSource object.

No Parameters
No Exceptions
createDOMWriter
Create a new DOMWriter object. DOMWriters are used to serialize a DOM tree back into an XML document.
Return Value

DOMWriter

The newly created DOMWriter object.

Note: By default, the newly created DOMWriter does not contain a DOMErrorHandler, i.e. the value of the errorHandler is null. However, implementations may provide a default error handler at creation time. In that case, the initial value of the errorHandler attribute on the new created DOMWriter contains a reference to the default error handler.

No Parameters
No Exceptions

1.3. Load Interfaces

A DOM application may use the hasFeature(feature, version) method of the DOMImplementation interface with parameter values "LS-Load" and "3.0" (respectively) to determine whether or not these interfaces are supported by the implementation. In order to fully support them, an implementation must also support the "Core" feature defined in the DOM Level 3 Core specification [DOM Level 3 Core].

A DOM application may use the hasFeature(feature, version) method of the DOMImplementation interface with parameter values "LS-Load-Async" and "3.0" (respectively) to determine whether or not the asynchronous mode is supported by the implementation. In order to fully support the asyncrhonous mode, an implementation must also support the "LS-Load" feature defined in this section.

Please, refer to additional information about conformance in the DOM Level 3 Core specification [DOM Level 3 Core].

Interface DOMBuilder

A interface to an object that is able to build a DOM tree from various input sources.

DOMBuilder provides an API for parsing XML documents and building the corresponding DOM document tree. A DOMBuilder instance is obtained by invoking the DOMImplementationLS.createDOMBuilder method.

As specified in [DOM Level 3 Core], when a document is first made available via the DOMBuilder:

  • there is only one Text node for each block of text. The Text nodes are in "normal" form: only structure (e.g. elements, comments, processing instructions, CDATA sections, and entity references) separates Text nodes, i.e., there are neither adjacent nor empty Text nodes.
  • it is expected that the value and nodeValue attributes of an Attr node initially return the XML 1.0 normalized value. However, if the boolean parameters validate-if-schema and datatype-normalization are set to true, depending on the attribute normalization used, the attribute values may differ from the ones obtained by the XML 1.0 attribute normalization. If the boolean parameter datatype-normalization is set to false, the XML 1.0 attribute normalization is guaranteed to occur, and if attributes list does not contain namespace declarations, the attributes attribute on Element node represents the property [attributes] defined in [XML Information set].
    Issue Infoset:
    XML Schemas does not modify the XML attribute normalization but represents their normalized value in an other information item property: [schema normalized value]
    Resolution: XML Schema normalization only occurs if datatype-normalization is set to true.

Asynchronous DOMBuilder objects are expected to also implement the events::EventTarget interface so that event listeners can be registered on asynchronous DOMBuilder objects.

Events supported by asynchronous DOMBuilder objects are:

  • load: The document that's being loaded is completely parsed, see the definition of LSLoadEvent
  • progress: Progress notification, see the definition of LSProgressEvent

Note: All events defined in this specification use the namespace URI "http://www.w3.org/2002/DOMLS".

Issue Parse-Security:
ED: State that a parse operation may fail due to security reasons (DOM LS telecon 20021202).

IDL Definition
interface DOMBuilder {
  readonly attribute DOMConfiguration config;
           attribute DOMBuilderFilter filter;
  readonly attribute boolean         async;
  readonly attribute boolean         busy;
  Document           parse(in DOMInputSource is)
                                        raises(DOMException);
  Document           parseURI(in DOMString uri)
                                        raises(DOMException);

  // ACTION_TYPES
  const unsigned short      ACTION_APPEND_AS_CHILDREN      = 1;
  const unsigned short      ACTION_REPLACE_CHILDREN        = 2;
  const unsigned short      ACTION_INSERT_BEFORE           = 3;
  const unsigned short      ACTION_INSERT_AFTER            = 4;
  const unsigned short      ACTION_REPLACE                 = 5;

  Node               parseWithContext(in DOMInputSource is, 
                                      in Node cnode, 
                                      in unsigned short action)
                                        raises(DOMException);
  void               abort();
};

Definition group ACTION_TYPES

A set of possible actions for the parseWithContext method.

Defined Constants
ACTION_APPEND_AS_CHILDREN
Append the result of the parse operation as children of the context node. For this action to work, the context node must be an Element or a DocumentFragment.
ACTION_INSERT_AFTER
Insert the result of the parse operation as the immediately following sibling of the context node. For this action to work the context node's parent must be an Element or a DocumentFragment.
ACTION_INSERT_BEFORE
Insert the result of the parse operation as the immediately preceeding sibling of the context node. For this action to work the context node's parent must be an Element or a DocumentFragment.
ACTION_REPLACE
Replace the context node with the result of the parse operation. For this action to work, the context node must have a parent, and the parent must be an Element or a DocumentFragment.
ACTION_REPLACE_CHILDREN
Replace all the children of the context node with the result of the parse operation. For this action to work, the context node must be an Element or a DocumentFragment.
Attributes
async of type boolean, readonly
True if the DOMBuider is asynchronous, false if it is synchronous.
busy of type boolean, readonly
True if the DOMBuider is currently busy loading a document, otherwise false.
config of type DOMConfiguration, readonly
The configuration used when a document is loaded. The values of parameters used to load a document are not passed automatically to the DOMConfiguration object used by the Document nodes. The DOM application is responsible for passing the parameters values from the DOMConfiguration object referenced from the DOMBuilder to the DOMConfiguration object referenced from the Document node.
In addition to the boolean parameters and parameters recognized in the Core module, the DOMConfiguration objects for DOMBuider adds the following boolean parameters:
"entity-resolver"
[required]
A DOMEntityResolver object. If this parameter has been specified, each time a reference to an external entity is encountered the implementation will pass the public and system IDs to the entity resolver, which can then specify the actual source of the entity.
If this parameter is not set, the resolution of entities in the document is implementation dependent.

Note: When the features "LS-Load" or "LS-Save" are supported, this parameter may also be supported by the DOMConfiguration object referenced from the Document node.

"certified"
true
[optional]
Assume, when XML 1.1 is supported, that the input is certified (see section 2.13 in [XML 1.1]).
false
[required] (default)
Don't assume that the input is certified (see section 2.13 in [XML 1.1]).
"charset-overrides-xml-encoding"
true
[required] (default)
If a higher level protocol such as HTTP [IETF RFC 2616] provides an indication of the character encoding of the input stream being processed, that will override any encoding specified in the XML declaration or the Text declaration (see also [XML 1.0] 4.3.3 "Character Encoding in Entities"). Explicitly setting an encoding in the DOMInputSource overrides encodings from the protocol.
false
[required]
Any character set encoding information from higher level protocols is ignored by the parser.
"supported-mediatypes-only"
true
[optional]
Check that the media type of the parsed resource is a supported media type. If an unsupported media type is encountered, a fatal error of type "unsupported-media-type" will be raised. The media types defined in [IETF RFC 3023] must always be accepted.
false
[required] (default)
Accept any media type.
"unknown-characters"
true
[required] (default)
If, while verifying full normalization when [XML 1.1] is supported, a processor encounters characters for which it cannot determine the normalization properties, then the processor will ignore any possible denormalizations caused by these characters.
This parameter is ignored [XML 1.0].
false
[optional]
Report an fatal error if a character is encountered for which the processor can not determine the normalization properties.
filter of type DOMBuilderFilter
When a filter is provided, the implementation will call out to the filter as it is constructing the DOM tree structure. The filter can choose to remove elements from the document being constructed, or to terminate the parsing early.
The filter is invoked after the operations requested by the DOMConfiguration parameters have been applied. For example, if "validate" is set to true, the validation is done before invoking the filter.
Methods
abort
Abort the loading of the document that is currently being loaded by the DOMBuilder. If the DOMBuilder is currently not busy, a call to this method does nothing.
No Parameters
No Return Value
No Exceptions
parse
Parse an XML document from a resource identified by a DOMInputSource.
Parameters
is of type DOMInputSource
The DOMInputSource from which the source of the document is to be read.
Return Value

Document

If the DOMBuilder is a synchronous DOMBuilder, the newly created and populated Document is returned. If the DOMBuilder is asynchronous, null is returned since the document object may not yet be constructed when this method returns.

Exceptions

DOMException

INVALID_STATE_ERR: Raised if the DOMBuilder's DOMBuilder.busy attribute is true.

parseURI
Parse an XML document from a location identified by a URI reference [IETF RFC 2396]. If the URI contains a fragment identifier (see section 4.1 in [IETF RFC 2396]), the behavior is not defined by this specification, future versions of this specification may define the behavior.
Parameters
uri of type DOMString
The location of the XML document to be read.
Return Value

Document

If the DOMBuilder is a synchronous DOMBuilder, the newly created and populated Document is returned. If the DOMBuilder is asynchronous, null is returned since the document object may not yet be constructed when this method returns.

Exceptions

DOMException

INVALID_STATE_ERR: Raised if the DOMBuilder's DOMBuilder.busy attribute is true.

parseWithContext
Parse an XML fragment from a resource identified by a DOMInputSource and insert the content into an existing document at the position specified with the contextNode and action arguments. When parsing the input stream, the context node is used for resolving unbound namespace prefixes. The context node's ownerDocument node is used to resolve default attributes and entity references.
As the new data is inserted into the document, at least one mutation event is fired per immediate child (or sibling) of context node.
If an error occurs while parsing, the caller is notified through the error handler.
Parameters
is of type DOMInputSource
The DOMInputSource from which the source document is to be read. The source document must be an XML fragment, i.e. anything except a complete XML document, a DOCTYPE (internal subset), entity declaration(s), notation declaration(s), or XML or text declaration(s).
cnode of type Node
The node that is used as the context for the data that is being parsed. This node must be a Document node, a DocumentFragment node, or a node of a type that is allowed as a child of an Element node, e.g. it can not be an Attribute node.
action of type unsigned short
This parameter describes which action should be taken between the new set of node being inserted and the existing children of the context node. The set of possible actions is defined above.
Return Value

Node

Return the node that is the result of the parse operation. If the result is more than one top-level node, the first one is returned.

Exceptions

DOMException

NOT_SUPPORTED_ERR: Raised if the DOMBuilder doesn't support this method.

NO_MODIFICATION_ALLOWED_ERR: Raised if the context node is readonly.

INVALID_STATE_ERR: Raised if the DOMBuilder's DOMBuilder.busy attribute is true.

Interface DOMInputSource

This interface represents a single input source for an XML entity.

This interface allows an application to encapsulate information about an input source in a single object, which may include a public identifier, a system identifier, a byte stream (possibly with a specified encoding), and/or a character stream.

The exact definitions of a byte stream and a character stream are binding dependent.

There are two places that the application will deliver this input source to the parser: as the argument to the parse method, or as the return value of the DOMEntityResolver.resolveEntity method.

(ED: There are at least three places where DOMInputSource is passed to the parser (parseWithContext).)

The DOMBuilder will use the DOMInputSource object to determine how to read XML input. If there is a character stream available, the parser will read that stream directly; if not, the parser will use a byte stream, if available; if neither a character stream nor a byte stream is available, the parser will attempt to open a URI connection to the resource identified by the system identifier.

A DOMInputSource object belongs to the application: the parser shall never modify it in any way (it may modify a copy if necessary).

Note: Even though all attributes in this interface are writable the DOM implementation is expected to never mutate a DOMInputSource.


IDL Definition
interface DOMInputSource {
           attribute DOMInputStream  byteStream;
  // The attribute characterStream is not available in ECMAScript
           attribute DOMReader       characterStream;
           attribute DOMString       stringData;
           attribute DOMString       encoding;
           attribute DOMString       publicId;
           attribute DOMString       systemId;
           attribute DOMString       baseURI;
};

Attributes
baseURI of type DOMString
The base URI to be used (see section 5.1.4 in [IETF RFC 2396]) for resolving relative URIs to absolute URIs. If the baseURI is itself a relative URI, the behavior is implementation dependent.
byteStream of type DOMInputStream
An attribute of a language-binding dependent type that represents a stream of bytes.
The parser will ignore this if there is also a character stream specified, but it will use a byte stream in preference to opening a URI connection itself.
If the application knows the character encoding of the byte stream, it should set the encoding attribute. Setting the encoding in this way will override any encoding specified in the XML declaration itself.
characterStream of type DOMReader, not available in ECMAScript
An attribute of a language-binding dependent type that represents a stream of 16-bit units. Application must encode the stream using UTF-16 (defined in [Unicode 2.0] and Amendment 1 of [ISO/IEC 10646]).
If a character stream is specified, the parser will ignore any byte stream and will not attempt to open a URI connection to the system identifier.
encoding of type DOMString
The character encoding, if known. The encoding must be a string acceptable for an XML encoding declaration ([XML 1.0] section 4.3.3 "Character Encoding in Entities").
This attribute has no effect when the application provides a character stream. For other sources of input, an encoding specified by means of this attribute will override any encoding specified in the XML declaration or the Text declaration, or an encoding obtained from a higher level protocol, such as HTTP [IETF RFC 2616].
publicId of type DOMString
The public identifier for this input source. The public identifier is always optional: if the application writer includes one, it will be provided as part of the location information.
stringData of type DOMString
A string attribute that represents a sequence of 16 bit units (utf-16 encoded characters).
If string data is available in the input source, the parser will ignore the character stream and the byte stream and will not attempt to open a URI connection to the system identifier.
systemId of type DOMString
The system identifier, a URI reference [IETF RFC 2396], for this input source. The system identifier is optional if there is a byte stream or a character stream, but it is still useful to provide one, since the application can use it to resolve relative URIs and can include it in error messages and warnings (the parser will attempt to fetch the ressource identifier by the URI reference only if there is no byte stream or character stream specified).
If the application knows the character encoding of the object pointed to by the system identifier, it can register the encoding by setting the encoding attribute.
If the system ID is a relative URI reference (see section 5 in [IETF RFC 2396]), the behavior is implementation dependent.
Interface DOMEntityResolver

DOMEntityResolver Provides a way for applications to redirect references to external entities.

Applications needing to implement customized handling for external entities must implement this interface and register their implementation by setting the entityResolver attribute of the DOMBuilder.

The DOMBuilder will then allow the application to intercept any external entities (including the external DTD subset and external parameter entities) before including them.

Many DOM applications will not need to implement this interface, but it will be especially useful for applications that build XML documents from databases or other specialized input sources, or for applications that use URNs.

Note: DOMEntityResolver is based on the SAX2 [SAX] EntityResolver interface.


IDL Definition
interface DOMEntityResolver {
  DOMInputSource     resolveEntity(in DOMString publicId, 
                                   in DOMString systemId, 
                                   in DOMString baseURI);
};

Methods
resolveEntity
Allow the application to resolve external entities.
The DOMBuilder will call this method before opening any external entity except the top-level document entity (including the external DTD subset, external entities referenced within the DTD, and external entities referenced within the document element); the application may request that the DOMBuilder resolve the entity itself, that it use an alternative URI, or that it use an entirely different input source.
Application writers can use this method to redirect external system identifiers to secure and/or local URIs, to look up public identifiers in a catalogue, or to read an entity from a database or other input source (including, for example, a dialog box).
If the system identifier is a URI, the DOMBuilder must resolve it fully before reporting it to the application through this interface.
(ED: See issue #4. An alternative would be to pass the URI out without resolving it, and to provide a base as an additional parameter. SAX resolves URIs first, and does not provide a base. )
Parameters
publicId of type DOMString
The public identifier of the external entity being referenced, or null if none was supplied.
systemId of type DOMString
The system identifier, a URI reference [IETF RFC 2396], of the external entity being referenced exactly as written in the source.
baseURI of type DOMString
The absolute base URI of the resource being parsed, or null if there is no base URI.
Return Value

DOMInputSource

A DOMInputSource object describing the new input source, or null to request that the parser open a regular URI connection to the system identifier.

No Exceptions
Interface DOMBuilderFilter

DOMBuilderFilters provide applications the ability to examine nodes as they are being constructed during a parse. As each node is examined, it may be modified or removed, or the entire parse may be terminated early.

At the time any of the filter methods are called by the parser, the owner Document and DOMImplementation objects exist and are accessible. The document element is never passed to the DOMBuilderFilter methods, i.e. it is not possible to filter out the document element. The Document, DocumentType, Notation, and Entity nodes are not passed to the filter.

All validity checking while reading a document occurs on the source document as it appears on the input stream, not on the DOM document as it is built in memory. With filters, the document in memory may be a subset of the document on the stream, and its validity may have been affected by the filtering.

All default content, including default attributes, must be passed to the filter methods.

Any exception raised in the filter are ignored by the DOMBuilder.

(ED: The description of these methods is not complete)

IDL Definition
interface DOMBuilderFilter {

  // Constants returned by startElement and acceptNode
  const short               FILTER_ACCEPT                  = 1;
  const short               FILTER_REJECT                  = 2;
  const short               FILTER_SKIP                    = 3;
  const short               FILTER_INTERRUPT               = 4;

  unsigned short     startElement(in Element elt);
  unsigned short     acceptNode(in Node enode);
  readonly attribute unsigned long   whatToShow;
};

Definition group Constants returned by startElement and acceptNode

Constants returned by startElement and acceptNode.

Defined Constants
FILTER_ACCEPT
Accept the node.
FILTER_INTERRUPT
Interrupt the normal processing of the document.
FILTER_REJECT
Reject the node abd its children.
FILTER_SKIP
Skip this single node. The children of this node will still be considered.
Attributes
whatToShow of type unsigned long, readonly
Tells the DOMBuilder what types of nodes to show to the filter. See NodeFilter for definition of the constants. The constant SHOW_ATTRIBUTE is meaningless here, attribute nodes will never be passed to a DOMBuilderFilter.
Methods
acceptNode
This method will be called by the parser at the completion of the parsing of each node. The node and all of its descendants will exist and be complete. The parent node will also exist, although it may be incomplete, i.e. it may have additional children that have not yet been parsed. Attribute nodes are never passed to this function.
From within this method, the new node may be freely modified - children may be added or removed, text nodes modified, etc. The state of the rest of the document outside this node is not defined, and the affect of any attempt to navigate to, or to modify any other part of the document is undefined.
For validating parsers, the checks are made on the original document, before any modification by the filter. No validity checks are made on any document modifications made by the filter.
If this new node is rejected, the parser might reuse the new node or any of its descendants.
Parameters
enode of type Node
The newly constructed element. At the time this method is called, the element is complete - it has all of its children (and their children, recursively) and attributes, and is attached as a child to its parent.
Return Value

unsigned short

  • FILTER_ACCEPT if this Node should be included in the DOM document being built.
  • FILTER_REJECT if the Node and all of its children should be rejected.
  • FILTER_SKIP if the Node should be skipped and the Node should be replaced by all the children of the Node.
  • FILTER_INTERRUPT if the filter wants to stop the processing of the document. Interrupting the processing of the document does no longer guarantee that the entire is XML well-formed.
No Exceptions
startElement
This method will be called by the parser after each Element start tag has been scanned, but before the remainder of the Element is processed. The intent is to allow the element, including any children, to be efficiently skipped. Note that only element nodes are passed to the startElement function.
The element node passed to startElement for filtering will include all of the Element's attributes, but none of the children nodes. The Element may not yet be in place in the document being constructed (it may not have a parent node.)
A startElement filter function may access or change the attributes for the Element. Changing Namespace declarations will have no effect on namespace resolution by the parser.
For efficiency, the Element node passed to the filter may not be the same one as is actually placed in the tree if the node is accepted. And the actual node (node object identity) may be reused during the process of reading in and filtering a document.
Parameters
elt of type Element
The newly encountered element. At the time this method is called, the element is incomplete - it will have its attributes, but no children.
Return Value

unsigned short

  • FILTER_ACCEPT if this Element should be included in the DOM document being built.
  • FILTER_REJECT if the Element and all of its children should be rejected. This return value will be ignored if elt is the documentElement, the documentElement can not be rejected.
  • FILTER_SKIP if the Element should be rejected. All of its children are inserted in place of the rejected Element node. This return value will be ignored if elt is the documentElement, the documentElement can not be rejected nor skipped.
  • FILTER_INTERRUPT if the filter wants to stop the processing of the document. Interrupting the processing of the document does no longer guarantee that the entire is XML well-formed.

Returning any other values will result in unspecified behavior.
No Exceptions
Interface LSProgressEvent

This interface represents a progress event object that notifies the application about progress as a document is parsed. It extends the Event interface defined in [DOM Level 3 Events].


IDL Definition
interface LSProgressEvent : events::Event {
  readonly attribute DOMInputSource  inputSource;
  readonly attribute unsigned long   position;
  readonly attribute unsigned long   totalSize;
};

Attributes
inputSource of type DOMInputSource, readonly
The input source that is being parsed.
position of type unsigned long, readonly
The current position in the input source, including all external entities and other resources that have been read.
totalSize of type unsigned long, readonly
The total size of the document including all external resources, this number might change as a document is being parsed if references to more external resources are seen.
Interface LSLoadEvent

This interface represents a load event object that signals the completion of a document load.


IDL Definition
interface LSLoadEvent : events::Event {
  readonly attribute Document        newDocument;
  readonly attribute DOMInputSource  inputSource;
};

Attributes
inputSource of type DOMInputSource, readonly
The input source that was parsed.
newDocument of type Document, readonly
The document that finished loading.

1.4. Save Interfaces

A DOM application may use the hasFeature(feature, version) method of the DOMImplementation interface with parameter values "LS-Save" and "3.0" (respectively) to determine whether or not these interfaces are supported by the implementation. In order to fully support them, an implementation must also support the "Core" feature defined in the DOM Level 3 Core specification [DOM Level 3 Core]. Please, refer to additional information about conformance in the DOM Level 3 Core specification [DOM Level 3 Core].

Interface DOMWriter

DOMWriter provides an API for serializing (writing) a DOM document out in an XML document. The XML data is written to an output stream, the type of which depends on the specific language bindings in use.

During serialization of XML data, namespace fixup is done when possible as defined in [DOM Level 3 Core], Appendix B. [DOM Level 2 Core] allows empty strings as a real namespace URI. If the namespaceURI of a Node is empty string, the serialization will treat them as null, ignoring the prefix if any.

Note: should the remark on DOM Level 2 namespace URI included in the namespace algorithm in Core instead?

DOMWriter accepts any node type for serialization. For nodes of type Document or Entity, well formed XML will be created if possible. The serialized output for these node types is either as a Document or an External Entity, respectively, and is acceptable input for an XML parser. For all other types of nodes the serialized form is not specified, but should be something useful to a human for debugging or diagnostic purposes. Note: rigorously designing an external (source) form for stand-alone node types that don't already have one defined in [XML 1.0] seems a bit much to take on here.

Within a Document, DocumentFragment, or Entity being serialized, Nodes are processed as follows

  • Document nodes are written including with the XML declaration and a DTD subset, if one exists in the DOM. Writing a Document node serializes the entire document.
  • Entity nodes, when written directly by DOMWriter.writeNode, output the entity expansion but no namespace fixup is done. The resulting output will be valid as an external entity.
  • EntityReference nodes are serialized as an entity reference of the form "&entityName;" in the output. Child nodes (the expansion) of the entity reference are ignored.
  • CDATA sections containing content characters that can not be represented in the specified output encoding are handled according to the "split-cdata-sections" boolean parameter.
    If the boolean parameter is true, CDATA sections are split, and the unrepresentable characters are serialized as numeric character references in ordinary content. The exact position and number of splits is not specified.
    If the boolean parameter is false, unrepresentable characters in a CDATA section are reported as errors. The error is not recoverable - there is no mechanism for supplying alternative characters and continuing with the serialization.
  • DocumentFragment nodes are serialized by serializing the children of the document fragment in the order they appear in the document fragment.
  • All other node types (Element, Text, etc.) are serialized to their corresponding XML source form.

Note: The serialization of a Node does not always generate a well-formed XML document, i.e. a DOMBuilder might through fatal errors when parsing the resulting serialization.

Within the character data of a document (outside of markup), any characters that cannot be represented directly are replaced with character references. Occurrences of '<' and '&' are replaced by the predefined entities &lt; and &amp;. The other predefined entities (&gt, &apos, and &quot;) are not used; these characters can be included directly. Any character that can not be represented directly in the output character encoding is serialized as a numeric character reference.

Attributes not containing quotes are serialized in quotes. Attributes containing quotes but no apostrophes are serialized in apostrophes (single quotes). Attributes containing both forms of quotes are serialized in quotes, with quotes within the value represented by the predefined entity &quot;. Any character that can not be represented directly in the output character encoding is serialized as a numeric character reference.

Within markup, but outside of attributes, any occurrence of a character that cannot be represented in the output character encoding is reported as an error. An example would be serializing the element <LaCañada/> with encoding="us-ascii".

When requested by setting the normalize-characters boolean parameter on DOMWriter, all data to be serialized, both markup and character data, is W3C Text normalized according to the rules defined in [CharModel]. The W3C Text normalization process affects only the data as it is being written; it does not alter the DOM's view of the document after serialization has completed.

Namespaces are fixed up during serialization, the serialization process will verify that namespace declarations, namespace prefixes and the namespace URIs associated with elements and attributes are consistent. If inconsistencies are found, the serialized form of the document will be altered to remove them. The method used for doing the namespace fixup while serializing a document is the algorithm defined in Appendix B.1 "Namespace normalization" of [DOM Level 3 Core].

(ED: previous paragraph to be defined closer here.)

Any changes made affect only the namespace prefixes and declarations appearing in the serialized data. The DOM's view of the document is not altered by the serialization operation, and does not reflect any changes made to namespace declarations or prefixes in the serialized output.

While serializing a document the serializer will write out non-specified values (such as attributes whose specified is false) if the discard-default-content boolean parameter is set to true. If the discard-default-content flag is set to false and a schema is used for validation, the schema will be also used to determine if a value is specified or not. If no schema is used, the specified flag on attribute nodes is used to determine if attribute values should be written out.

Ref to Core spec (1.1.9, XML namespaces, 5th paragraph) entity ref description about warning about unbound entity refs. Entity refs are always serialized as &foo;, also mention this in the load part of this spec.


IDL Definition
interface DOMWriter {
  readonly attribute DOMConfiguration config;
           attribute DOMString       encoding;
           attribute DOMString       newLine;
           attribute DOMWriterFilter filter;
  boolean            writeNode(in DOMOutputStream destination, 
                               in Node wnode);
  DOMString          writeToString(in Node wnode)
                                        raises(DOMException);
};

Attributes
config of type DOMConfiguration, readonly
The configuration used when a document is loaded. The values of parameters used on a document are not passed automatically from the DOMConfiguration object used by the Document nodes. The DOM application is responsible for passing the parameters values from the DOMConfiguration object referenced from the Document node to the DOMConfiguration object referenced from the DOMWriter.
In addition to the boolean parameters and parameters recognized in the Core module, the DOMConfiguration objects for DOMWriter adds, or modifies, the following boolean parameters:
"entity-resolver"
This parameter is equivalent to the "entity-resolver" parameter defined in DOMBuilder.config.
"xml-declaration"
true
[required] (default)
If a Document Node or an Entity node is serialized, the XML declaration, or text declaration, should be included Document.version and/or an encoding is specified.
false
[required]
Do not serialize the XML and text declarations.
"canonical-form"
true
[optional]
This formatting writes the document according to the rules specified in [Canonical XML]. Setting this boolean parameter to true will set the boolean parameter "format-pretty-print" to false.
false
[required] (default)
Do not canonicalize the output.
"format-pretty-print"
true
[optional]
Formatting the output by adding whitespace to produce a pretty-printed, indented, human-readable form. The exact form of the transformations is not specified by this specification. Setting this boolean parameter to true will set the boolean parameter "canonical-form" to false.
false
[required] (default)
Don't pretty-print the result.
"normalize-characters"
This boolean parameter is equivalent to the one defined by DOMConfiguration in [DOM Level 3 Core]. Unlike in the Core, the default value for this boolean parameter is true. While DOM implementations are not required to implement the W3C Text Normalization defined in [CharModel], this boolean parameter must be activated by default if supported.
"unknown-characters"
true
[required] (default)
If, while verifying full normalization when [XML 1.1] is supported, a character is encountered for which the normalization properties cannot be determined, then ignore any possible denormalizations caused by these characters.
false
[optional]
Report an fatal error if a character is encountered for which the processor can not determine the normalization properties.
encoding of type DOMString
The character encoding in which the output will be written.
The encoding to use when writing is determined as follows:
  • If the encoding attribute has been set, that value will be used.
  • If the encoding attribute is null or empty, but the item to be written, or the owner document of the item, specifies an encoding (i.e. the "actualEncoding" from the document) specified encoding, that value will be used.
  • If neither of the above provides an encoding name, a default encoding of "UTF-8" will be used.

The default value is null.
filter of type DOMWriterFilter
When the application provides a filter, the serializer will call out to the filter before serializing each Node. Attribute nodes are never passed to the filter. The filter implementation can choose to remove the node from the stream or to terminate the serialization early.
newLine of type DOMString
The end-of-line sequence of characters to be used in the XML being written out. Any string is supported, but these are the recommended end-of-line sequences (using other character sequences than these recommended ones can result in a document that is either not serializable or not well-formed):
null
Use a default end-of-line sequence. DOM implementations should choose the default to match the usual convention for text files in the environment being used. Implementations must choose a default sequence that matches one of those allowed by "End-of-Line Handling" ([XML 1.0], section 2.11) if the serialized content is XML 1.0 or "End-of-Line Handling" ([XML 1.1], section 2.11) if the serialized content is XML 1.1.
CR
The carriage-return character (#xD).
CR-LF
The carriage-return and line-feed characters (#xD #xA).
LF
The line-feed character (#xA).

The default value for this attribute is null.
Methods
writeNode
Write out the specified node as described above in the description of DOMWriter. Writing a Document or Entity node produces a serialized form that is well formed XML, when possible (Entity nodes might not always be well formed XML in themselves). Writing other node types produces a fragment of text in a form that is not fully defined by this document, but that should be useful to a human for debugging or diagnostic purposes.
If the specified encoding is not supported the error handler is called and the serialization is interrupted.
Parameters
destination of type DOMOutputStream
The destination for the data to be written.
wnode of type Node
The Document or Entity node to be written. For other node types, something sensible should be written, but the exact serialized form is not specified.
Return Value

boolean

Returns true if node was successfully serialized and false in case a failure occured and the failure wasn't canceled by the error handler.

No Exceptions
writeToString
Serialize the specified node as described above in the description of DOMWriter. The result of serializing the node is returned as a DOMString (this method completely ignores all the encoding information available). Writing a Document or Entity node produces a serialized form that is well formed XML. Writing other node types produces a fragment of text in a form that is not fully defined by this document, but that should be useful to a human for debugging or diagnostic purposes.
Error handler is called if encoding not supported...
Parameters
wnode of type Node
The node to be written.
Return Value

DOMString

Returns the serialized data, or null in case a failure occured and the failure wasn't canceled by the error handler.

Exceptions

DOMException

DOMSTRING_SIZE_ERR: Raised if the resulting string is too long to fit in a DOMString.

Interface DOMWriterFilter

DOMWriterFilters provide applications the ability to examine nodes as they are being serialized. DOMWriterFilter lets the application decide what nodes should be serialized or not.

The Document, DocumentType, Notation, and Entity nodes are not passed to the filter.


IDL Definition
interface DOMWriterFilter : traversal::NodeFilter {
  readonly attribute unsigned long   whatToShow;
};

Attributes
whatToShow of type unsigned long, readonly
Tells the DOMWriter what types of nodes to show to the filter. See NodeFilter for definition of the constants. The constants SHOW_ATTRIBUTE, SHOW_DOCUMENT, SHOW_DOCUMENT_TYPE, SHOW_NOTATION, and SHOW_DOCUMENT_FRAGMENT are meaningless here, those nodes will never be passed to a DOMWriterFilter. Entity nodes are not passed to the filter.

1.5. Convenience Interfaces

The interfaces defined in this section provide no direct functionality that can not be achieved with the load and save interfaces defined in the earlier sections of this specification. These interfaces are defined for developer convenience only, and supporting them is optional.

Interface DocumentLS

The DocumentLS interface provides a mechanism by which the content of a document can be serialized, or replaced with the DOM tree produced when loading a URI, or parsing a string.

If the DocumentLS interface is supported, the expectation is that an instance of the DocumentLS interface can be obtained by using binding-specific casting methods on an instance of the Document interface, or by using the method Node.getFeature with parameter values "LS-Load" and "3.0" (respectively) on an Document, if the Document supports the feature "Core" version "3.0" defined in [DOM Level 3 Core]

This interface is optional. If supported, implementations are must support version "3.0" of the feature "LS-DocumentLS".


IDL Definition
interface DocumentLS {
           attribute boolean         async;
                                        // raises(DOMException) on setting

  void               abort();
  boolean            load(in DOMString uri);
  boolean            loadXML(in DOMString source);
  DOMString          saveXML(in Node snode)
                                        raises(DOMException);
};

Attributes
async of type boolean
Indicates whether the method DocumentLS.load() should be synchronous or asynchronous. When the async attribute is set to true the load method returns control to the caller before the document has completed loading. The default value of this attribute is true.
Issue async-1:
Should the DOM spec define the default value of this attribute? What if implementing both async and sync IO is impractical in some systems?
Resolution: 2001-09-14. default is false but we need to check with Mozilla and IE.
2003-01-24. Checked with IE and Mozilla, default is true.
Exceptions on setting

DOMException

NOT_SUPPORTED_ERR: Raised if the implementation doesn't support the mode the attribute is being set to.

Methods
abort
If the document is currently being loaded as a result of the method load being invoked the loading and parsing is immediately aborted. The possibly partial result of parsing the document is discarded and the document is cleared.
No Parameters
No Return Value
No Exceptions
load
Replaces the content of the document with the result of parsing the given URI. Invoking this method will either block the caller or return to the caller immediately depending on the value of the async attribute. Once the document is fully loaded the document will fire a "load" event that the caller can register as a listener for. If an error occurs the document will fire an "error" event so that the caller knows that the load failed (see ParseErrorEvent). If this method is called on a document that is currently loading, the current load is interrupted and the new URI load is initiated.
When invoking this method the features used in the DOMBuilder interface are assumed to have their default values with the exception that the feature "entities" is "false".
Parameters
uri of type DOMString
The URI reference for the XML file to be loaded. If this is a relative URI, the base URI used by the implementation is implementation dependent.
Return Value

boolean

If async is set to true load returns true if the document load was successfully initiated. If an error occurred when initiating the document load load returns false.
If async is set to false load returns true if the document was successfully loaded and parsed. If an error occurred when either loading or parsing the URI load returns false.

No Exceptions
loadXML
Replace the content of the document with the result of parsing the input string, this method is always synchronous. This method always parses from a DOMString, which means the data is always UTF16. All other encoding information is ignored.
The features used in the DOMBuilder interface are assumed to have their default values when invoking this method.
Parameters
source of type DOMString
A string containing an XML document.
Return Value

boolean

true if parsing the input string succeeded without errors, otherwise false.

No Exceptions
saveXML
Save the document or the given node and all its descendants to a string (i.e. serialize the document or node).
The features used in the DOMWriter interface are assumed to have their default values when invoking this method.
Parameters
snode of type Node
Specifies what to serialize, if this parameter is null the whole document is serialized, if it's non-null the given node is serialized.
Return Value

DOMString

The serialized document or null in case an error occured.

Exceptions

DOMException

WRONG_DOCUMENT_ERR: Raised if the node passed in as the node parameter is from an other document.

Interface ElementLS

The ElementLS interface provides a convenient mechanism by which the children of an element can be serialized to a string, or replaced by the result of parsing a provided string.

If the ElementLS interface is supported, the expectation is that an instance of the ElementLS interface can be obtained by using binding-specific casting methods on an instance of the Element interface, or by using the method Node.getFeature with parameter values "LS-Load" and "3.0" (respectively) on an Element, if the Element supports the feature "Core" version "3.0" defined in [DOM Level 3 Core].

This interface is optional. If supported, implementations must support version "3.0" of the feature "LS-ElementLS".


IDL Definition
interface ElementLS {
           attribute DOMString       markupContent;
};

Attributes
markupContent of type DOMString
The content of the element in serialized form.
When getting the value of this attribute, the children are serialized in document order and the serialized result is returned. This is equivalent of calling DOMWriter.writeToString() on all children in document order and appending the result of the individual results to a single string that is then returned as the value of this attribute.
When setting the value of this attribute, all children of the element are removed, the provided string is parsed and the result of the parse operation is inserted into the element. This is equivalent of calling DOMBuilder.parseWithContext() passing in the provided string (through the input source argument), the Element, and the action ACTION_REPLACE_CHILDREN. If an error occures while parsing the provided string, the Element's owner document's error handler will be called, and the Element is left with no children.
Both setting and getting the value of this attribute assumes that the parameters in the DOMConfiguration object have their default values.

1.6. Issue List

1.6.1. Open Issues

Issue LS-Issue-90:
The interaction and relationships between all the DOMBuilder and DOMWriter features need to be defined, i.e. setting x will set y and unset z.
Issue LS-Issue-15:
System Exceptions. Loading involves file opens and reads, and these can result in a variety of system errors that may already have associated system exceptions. Should these system exceptions pass through as is, or should they be some how wrapped in DOMExceptions, or should there be a parallel set DOM Exceptions, or what?
Issue LS-Issue-21:
Define exceptions. A DOMSystemException needs to be defined as part of the error handling module that is to be shared with AS. Common I/O type errors need to be defined for it, so that they can be reported in a uniform way. A way to embed errors or exceptions from the OS or language environment is needed, to provide full information to applications that want it.
Issue LS-Issue-58:
Some features should not be required for parseWithContext() (such as validate, validate-if-schema, whitespace-in-element-content, external-dtd-subset, ...), what are these options, and how do we describe this?
Issue LS-Issue-108:
Do we want to have NodeLS.load()/loadXML()? See

1.6.2. Resolved Issues

Issue LS-Issue-1:
Should these methods be in a new interface, or should they be added to the existing DOMImplementation Interface? I think that adding them to the existing interface is cleaner, because it helps avoid an explosion of new interfaces.
The methods are in a separate interface in this description for convenience in preparing the doc, so that I don't need to edit Core to add the methods. (The same argument could perhaps be made for implementations.)
Resolution: The methods are in a separate DOMImplementationLS interface. Because Load/Save is an optional module, we don't want to add its to the core DOMImplementation interface.
Issue LS-Issue-2:
SAX handles the setting of parser attributes differently. Rather than having distinct getters and setters for each attribute, it has a generic setter and getter of named properties, where properties are specified by a URI. This has an advantage in that implementations do not need to extend the interface when providing additional attributes.
If we choose to use strings, their syntax needs to be chosen. URIs would make sense, except for the fact that these are just names that do not refer to any resources. Dereferencing them would be meaningless. Yet the direction of the W3C is that all URIs must be dereferencable, and refer to something on the web.
Resolution: Use strings for properties. Use Java package name syntax for the identifying names. The question was revisited at the July f2f, with the same conclusion. But some discussion of using URIs continues.
This issue was revisited once again at the 9/2000 meeting. Now all DOM properties or features will be short, descriptive names, and we will recommend that all vendor-specific extensions be prefixed to avoid collisions, but will not make specific recommendations for the syntax of the prefix.
Issue LS-Issue-3:
It's not obvious what name to choose for the parser interface. Taking any of the names already in use by parser implementations would create problems when trying to support both the new API and the existing old API. That leaves out DocumentBuilder (Sun) and DOMParser (Xerces).
Resolution: This is issue really just a comment. The "resolution" is in the names appearing in the API.
Issue LS-Issue-4:
Question: should ResolveEntity pass a baseURI string back to the application, in addition to the publicId, systemId, and/or stream? Particularly in the case of an input stream.
Resolution: No. Sax2 explicitly says that the system ID URI must be fully resolved before passing it out to the entity resolve. We will follow SAX's lead on this unless some additional use case surfaces. This is from the 9/2000 f2f, and reverses an earlier decision.
2002-02-22: a baseURI parameter was added.
Issue LS-Issue-5:
When parsing a document that contains errors, should the whole document be decreed unusable, or should we say that portions prior to the point where the error was detected are OK?
Resolution: In the case of errors in the XML source, what, if any, document is returned is implementation dependent.
Issue LS-Issue-6:
The relationship between SAXExceptions and DOM exceptions seems confusing.
Resolution: This issue goes away because we are no longer using SAX. Any exceptions will be DOM Exceptions.
Issue LS-Issue-7:
Question: In the original Java definition, are the strings returned from the methods SAXException.toString and SAXException.getMessage always the same? If not, we need to add another attribute.
Resolution: No longer an issue because we are no longer using SAX.
Issue LS-Issue-8:
JAXP defines a mechanism, based on Java system properties, by which the Document Builder Factory locates the specific parser implementation to be used. This ability to redirect to different parsers is a key feature of JAXP. How this redirection works in the context of this design may be something that needs to be defined separately for each language binding.
This question was discussed at the July f2f, without resolution. Agreed that the feature is not critical to the rest of the API, and can be postponed.
Resolution: The issue is moving to core, where it is part of the bigger question of where does the DOM implementation come from, and how do multiple implementations coexist. Allowing separate, or mix-and-match, specification of the parser and the rest of the DOM is not generally practical because parsers generally have some degree of private knowledge about their DOMs.
Issue LS-Issue-9:
The use of interfaces from SAX2 raises some questions. The Java bindings for these interfaces need to be exactly the SAX2 definitions, including the original org.xml.sax package name.
The IDL presented here for these interfaces is an attempt to map the Java into IDL, but it will certainly not round-trip accurately - Java bindings generated from the IDL will not match the original Java.
The reasons for using the SAX interfaces are that they are well designed, widely implemented and used, and provide what is needed. Designing something new would create confusion for application developers (which should be used?) and make extra work for implementers of the DOM, most of whom probably already provide SAX, all for no real gain.
Resolution: Problem is gone. We are not using SAX2. The design will borrow features and concepts from SAX2 when it makes sense to do so.
Issue LS-Issue-10:
Error Reporting. Loading will be reporting well-formedness and validation errors, just like AS. A common error reporting mechanism needs to be developed.
Resolution: Resolved, see errors.html
Issue LS-Issue-11:
Another Error Reporting Question. We decided at the June f2f that validity errors should not be exceptions. This means that a document load operation could encounter multiple errors. Should these be collected and delivered as some sort of collection at the (otherwise) successful completion of the load, or should there be some sort of callback? Callbacks are harder for applications to deal with.
Resolution: Provide a callback mechanism. Provide a default error handler that throws an exception and stops further processing. From July f2f.
Issue LS-Issue-12:
Definition of "Non-validating". Exactly how much processing is done by "non-validating" parsers is not fully defined by the XML specification. In particular, they are not required to read any external entities, but are not prohibited from doing so.
Another common user request: a mode that completely ignores DTDs, both and external. Such a parser would not conform to XML 1.0, however.
For the documents produced by a non-validating load to be the same, we need to tie down exactly what processing must be done. The XML Core WG also has question as an open issue .
Some discussion is at http://lists.w3.org/Archives/Member/w3c-xml-core-wg/2000JanMar/0192.html
Here is proposal: Have three classes of parsers
  • Minimal. No external entities of any type are accessed. DTD subset is processes normally, as required by XML 1.0, including all entity definitions it contains.
  • Non-Validating. All external entities are read. Does everything except validation.
  • Validating. As defined by XML 1.0 rec.

Resolution: Use the options from SAX2. These provide separate flags for validation, reading of external general entities and reading of external parameter entities.
Issue LS-Issue-13:
Use of System or Language specific types for Input and Output
Loading and Saving requires that one of the possible sources or destinations of the XML data be some sort of stream that can be used with io streams or memory buffers, or anything else that might take or supply data. The type will vary, depending on the language binding.
The question is, what should be put into the IDL interfaces for these? Should we define an XML stream to abstract out the dependency, or use system classes directly in the bindings?
Resolution: Define IDL types for use in the rest of the interface definitions. These types will be mapped directly to system types for each language binding
Issue LS-Issue-14:
Should there be separate DOM modules for browser or scripting style loading (document.load("whatever")) and server style parsers? It's probably easy for the server style parsers to implement the browser style interface, but the reverse may not be true.
Resolution: Yes. A client application style API will be provided.
Issue LS-Issue-16:
Loading and saving of abstract schema's - DTDs or Schemas - outside of the context of a document is not addressed.
Resolution: See the DOMASBuilder interface in the AS spec
Issue LS-Issue-17:
Loading while validating using an already loaded abstract schema is not addressed. Applications should be able to load a abstract schema (issue 16), and then repeatedly reuse it during the loading of additional documents.
Resolution: See the DOMASBuilder interface in the AS spec
Issue LS-Issue-18:
For the list of parser properties, which must all implementations recognize, which settings must all implementations support, and which are optional?
Resolution: Done
Issue LS-Issue-19:
DOMOutputStream: should this be an interface with methods, or just an opaque type that maps onto an appropriate binding-specific stream type?
If we specify an actual interface with methods, applications can implement it to wrap any arbitrary destination that they may have. If we go with the system type it's simpler to output to that type of stream, but harder otherwise.
Resolution: Opaque.
Issue LS-Issue-20:
Action from September f2f to "add issues raised by schema discussion". What were these?
Resolution: nobody seems to remember this, no action taken
Issue LS-Issue-22:
What do the bindings for things like InputStream look like in ECMA Script? Tentative resolution - InputStream will map to a binding dependent class or interface. For environments where nothing appropriate exists, a new interface will be created. This question is still being discussed.
Resolution: will be left to the binding
Issue LS-Issue-23:
To Do: Add a method or methods to DOMBuilder that will provide information about a parser feature - is the name recognized, which (boolean) values are supported - without throwing exceptions.
Resolution: Done. Added canSetFeature.
Issue LS-Issue-24:
Clearly identify which of the parser properties must be recognized, and which of their settings must be supported by all conforming implementations.
Resolution: Done. All must be recognized.
Issue LS-Issue-25:
How does the validation property work in SAX, and how should it work for us? The default value in SAX2 is "true". Non-validating parsers only support a value of false. Does this mean that the default depends on the parser, or that some sort of an error happens if a parse is attempted before resetting the property, or what?
The same question applies to the External Entities properties too.
Resolution: Make the default value for the validation property be false.
Issue LS-Issue-26:
Do we want to rename the "auto-validation" property to "validate-if-cm"? Proposed at f2f. Resolution unclear.
Resolution: Changed the name to "validate-if-cm".
Issue LS-Issue-27:
How is validation during document loading handled when there are multiple possible abstract schemas associated with the document? How is one selected? The same question exists for documents in general, outside of the context of loading. Resolving the question for loading probably needs to wait until the more general question is understood.
Resolution: Always use the active external AS if any and the active internal AS if any. Whenever you want to validate during parsing with a different Internal/External model you have to activate this Abstract Schema first.
Issue LS-Issue-29:
Should all properties except namespaces default to false? Discussed at f2f. I'm not so sure now. Some of the properties have somewhat non-standard behavior when false - leaving out ER nodes or whitespace, for example - and support of false will probably not even be required.
Resolution: Not all properties should default to false. But validation should.
Issue LS-Issue-28:
To do: add new parser property "createEntityNodes". default is true. Illegal for it to be false and createEntityReferenceNodes to be true.
(ED: Is this really what we want? )

Resolution: new feature added.
Issue LS-Issue-30:
Possible additional parser features - option to not create CDATA nodes, and to merge CDATA contents with adjacent TEXT nodes if they exist. Otherwise just create a TEXT node.
Option to omit Comments.
Resolution: new feature added.
Issue LS-Issue-31:
We now have an option for fixing up namespace declarations and prefixes on serialization. Should we specify how this is done, so that the documents from different implementations of serialization will use the same declarations and prefixes, or should we leave the details up to the implementation?
Resolution: The exact form of the namespace fixup is implementation dependent. The only requirement is that all elements and attributes end up with the correct namespace URI.
Issue LS-Issue-32:
Mimetypes. If the input being parsed is from http or something else that supplies types, and the type is something other than text/xml, should we parse it anyhow, or should we complain. Should there be an option?
Tentative resolution: always parse, never complain. Reasons: 1. This is what all parsers do now, and no one has ever complained, at least not that I'm aware of. 2. Applications must have a pretty good reason to suspect that they're getting xml or they wouldn't have invoked the parser. 3. All the test would do is to take something that might have worked (xml that is not known to the server) and turn it into an error. Non-xml is exceptionally unlikely to successfully parse (be well formed.)
Resolution: See the supported-mediatypes-only feature on DOMBuilder.
Issue LS-Issue-33:
Unicode Character Normalization Problems. It turns out that for some code pages, normalizing a Unicode representation, translating to the code page, then translating back to Unicode can result in un-normalized Unicode. Mark Davis says that this can happen with Vietnamese and maybe with Hebrew.
This means that the suggested W3C model of normalization on serialization (early normalization) may not work, and that the receiver of the data may need to normalize it again, just in case.
Resolution: The scenario described is a quality-of-implementation issue. A transcoder converting from the one of the troublesome code pages to a Unicode representation should be responsible for re-normalizing the output.
Issue LS-Issue-34:
Features 2.1.4.1, 2 - XML Fragment Support. Should these be dropped?
Resolution: The DOM WG decided to drop support for XML fragment loading in the DOM Level 3 Load-Save module due to lack of time to define the behavior in all the edge cases, future versions of this spec might address this issue.
Issue LS-Issue-35:
XPath based document load filter. It would be plausible to have a partial (filtered) document load based on selecting the portion of the document to load with an XPath expression. This facility could be in addition to the node-by-node filtering currently specified. Or we could drop the existing filter.
Implementing an XPath based selective load would require that there be an XPath processor present in addition to the parser itself.
Resolution: The DOM Level 3 spec will not define an interface for doing XPath/XPointer type filtering, implementations are free to implement XPath/XPointer based filters on top of a DOMBuilderFilter.
Issue LS-Issue-36:
MIME Type checking for DOMASBuilder.
What MIME Type checking needs to be done for parsing schemas
Resolution: see DOMBuilder, DOMASBuilder is an extend of DOMBuilder, this issue is solved within DOMBuilder
Issue LS-Issue-37:
Internal ASModel serialization for DOMWriter.
What if the internal ASModel is an XML Schema ASModel. Currently there is no ASModel type. Adding an Internal ASModel can be any kind of schema. Should serialization somehow check the internal ASModel ? What about the internal subset, is it discarded when the AS spec is implemented ?
Resolution: An internal ASModel can't be a schema according to the AS spec. The internal subset is discarded when an Abstract Schema is active and the AS spec is implemented
Issue LS-Issue-38:
Attribute Normalization.
Add a property to "attributeNormalization" to DOMWriter to support or discard Attribute Normalization during serialization to. Setting attributeNormalization will serialize attributes with unexpanded entity references (if any) regardless their childnode(s). This means that if a user is changing the child nodes of an entity reference node within an attribute and attributeNormalization is set to true during serialization that these changes are discarded during serialization.
Resolution: The normalization will be driven by the validation options on DOMBuilder, if a document is validated it will also be normalized, if the document is not validated then no normalization will occure.
Issue LS-Issue-39:
Validation at serialization time. Should we have an option for validating while serializing, what about validation errors, should we allow serializing non-valid DOM's?
Resolution: No. Validation at serialization time will not be supported by this specification.
Issue LS-Issue-40:
Is the description of the DOMWriter option expand-entity-references acceptable?
Resolution: Yes, the description is acceptable.
Issue LS-Issue-41:
Do we need filter support in DOMWriter too?
Resolution: Not until we have good usecases for needing filters when serializing a node.
Issue LS-Issue-42:
Should all attributes on DOMInputSource be readonly? The DOM implementation will be passed an object that implements this interface and there's no need for the DOM implementation to ever modify any of those values.
Resolution: Yes, the application is responsible for implementing this interface, the DOM implementation should never modify an input source.
Issue LS-Issue-43:
What's a DOMReader in non-Java languages? Does this really belong in these language neutral interfaces?
Resolution: The DOMReader type should be defined as "Object" in ECMAScript.
Issue LS-Issue-44:
What should the DOMWriter do if the doctype name doesn't match the name of the document element? This is a validity error, not a wellformedness error so should this just be a normal validity error when serializing?
Resolution: This is only a validity error, and since this spec doesn't support validation at serialization time this will be ignored. If an implementation were to support validation at serialization time the error handler should be called in this case.
Issue LS-Issue-45:
How should validation work if there's a reference to both a schema and a DTD, should the parser validate against both, or only one, if only one, how does one select which one?
Resolution: Add a validate-against-dtd option that forces validation against the DTD even if there are other schemas referenced in the document.
Issue LS-Issue-46:
Should supporting async/sync loading be optional?
Resolution: Yes.
Issue LS-Issue-47:
Default attribute handling in DOMWriter needs to be defined for Level 1 elements.
Resolution: If Attr.specified is set to false then the attribute must be a level 1 node in which case this information can safely be used.
(ED: This resolution needs to be put in sync with our Attr.specified discussion.)
Issue LS-Issue-48:
DOMWriter::writeNode takes a Node as an argument, shouldn't this be a Document?
Resolution: It should also be possible to serialize elements, adding xmlns declarations on the element that is serialized. Entities get serialized w/o binding element namespaces. Text nodes should be serialized too, and document fragments, cdata section and attributes too and entity reference (&foo;) and comments.
Issue LS-Issue-49:
Datatype normalization? I.e. stripping whitespace around integers n' such.
Resolution: No, but add option to not normalize when validating, "datatype-normalization" added.
Issue LS-Issue-50:
Should 'external-parameter-entities' be replaced by an "load-external-dtds-n'-stuff" option?
Resolution: yes, done, "external-parameter-entities" added.
Issue LS-Issue-51:
DOMBuilder::canSetFeature and ::supportsFeature are redundant, no?
Resolution: Yes, supportsFeature removed.
Issue LS-Issue-52:
Is the API dependencies on the Events spec acceptable?
Resolution: We're only reusing events API's, we're not requiring people to implement the events spec so this shouldn't be a problem.
Issue LS-ISSUE-53:
Doesn't the feature "external-dtd-subset" conflict with the XML 1.0 specifications standalone="true"?
Resolution: No, the standalone "attribute" in XML 1.0 is only a hint, and thus implementations are not required to do anything with it that matters for a DOM builder.
Issue LS-Issue-54:
"canonical-form" needs a correct reference to the spec for canonical XML.
Issue LS-Issue-55:
How should default attributes be dealt with wrt DOMBuilderFilter?
Resolution: All default content must be passed to the filter.
Issue LS-Issue-56:
Should we make it possible to SKIP an element in DOMBuilderFilter::acceptNode?
Resolution: Yes, done.
Issue LS-Issue-57:
namespaceURI in core can be empty string, how should that be dealt with in DOM LS?
Resolution: [DOM Level 2 Core] allows empty strings as a real namespace URI. If the namespaceURI of a Node is empty string, the serialization will treat them as null, ignoring the prefix if any.
Issue LS-Issue-59:
ACTION_APPEND is confusing, can we clarify it?
Resolution: make it ACTION_APPEND_AS_CHILDREN (2002-01-28)
Issue LS-Issue-60:
DOMEntityResolver::baseURI, should it be absolute or can it be relative?
Resolution: make it absolute. (2002-01-28)
Issue LS-Issue-61:
How to use an empty document with parseWithContext?
Resolution: As of today, it is not possible to have an empty Document using the DOM Core, so we don't consider this as an issue. However, following the discussion on having support for empty Document in the Core, this issue might be reopened.
Issue LS-Issue-62:
createDOMBuilder: If MODE_SYNCHRONOUS and MODE_ASYNCHRONOUS are the only anticipated values, then a boolean parameter would be preferred. If it stays a unsigned short, then there needs to be a exception for unrecognized values.
Resolution: We keep the unsigned hsort for future possible extension.
"NOT_SUPPORTED_ERR: Raised if the requested mode is not supported."
Issue LS-Issue-63:
createDOMBuilder: The description of the return value mentions the type parameter, however the method has no parameters.
Resolution: Fixed.
Issue LS-Issue-64:
createDOMWriter: Being able to create an asynchronous writer would be desirable. I'd add a mode parameter to parallel createDOMBuilder.
Resolution: This will not be addressed by this version of the DOM LS spec.
Issue LS-Issue-65:
DOMBuilder.errorHandler: Passing "the node closest to where the error occurred" is really vague. Especially if the problem is a well-formedness or other fatal error. An character offset and/or text fragment would be more useful for error diagnosis. Passing null if the closest node could not be determined would be cleaner than passing the document.
Resolution: Description updated to indicate that any other available position information should also be passed to the error handler.
Issue LS-Issue-66:
parse and parseURI DOMBuilder methods: Returning null for asynchrous DOMBuilder's would make it difficult to express DocumentLS.load in terms of DOMBuilder.parse. Since DocumentLS appears to be a convenience interface, everything should be expressible in terms of the more general interfaces.
Resolution: DocumentLS.load and DOMBuilder.parse* are two completely different animals. One can most likely not be implemented in terms of using the other, and this will not change. DocumentLS.load is defined as it is for compatibility with existing implementations, and that won't be changed. Returning a document from an async parse method on the DOMBuilder is just not practical since you don't know at the time when the parse method returns what type of document you'll need. No change.
Issue LS-Issue-67:
DOMBuilder.parseURI: Specifying a behavior for URI's containing fragment identifier would seem desirable. I'd suggest ignoring the fragment identifier, but throwing an exception would be better than leaving it unspecified.
Resolution: Description updated, no exception, undefined behavior for now but future versions might define the behavior.
Issue LS-Issue-68:
DOMBuilder.parseWithContext: Should throw DOMSystemExceptions. Should throw NO_MODIFICATION_ALLOWED_ERR if context node (or parent) is read-only. Returning the created node would be desirable.
Resolution: Exception added. But the created node can't be returned since there might be more than one node created.
Issue LS-Issue-69:
How does DOMBuilder.parseWithContext interact with any event listeners registered on the context node or its ancestors?
Resolution: Description on what mutation events are fired when using parseWithContext() added.
Issue LS-Issue-70:
DOMBuilder.setFeature: Several features force other features to specific values, but there is no defined behavior if you try to override the forced value, for example, setting external-parameter-entities to false after setting validation to true. I would suggest throwing an exception.
Resolution: No exceptions will be thrown. See issue 90.
Issue LS-Issue-71:
DOMWriter.encoding attribute: The second bullet should describe how Document.encoding or Document.actualEncoding are used to determine the encoding.
Issue LS-Issue-72:
DOMWriter.encoding attribute: Should throw an exception on setting if the encoding in not supported.
Resolution: Definition of DOMWriter.writeNode() updated, no exception thrown on setting the encoding.
Issue LS-Issue-73:
DOMWriter.encoding attribute: There should be a list of required encodings (at minimum UTF-8 and UTF-16)
Resolution: No list will be defined in the DOM spec. The XML specification defines some required encodings, we won't define anything more than that.
Issue LS-Issue-74:
DOMWriter.lastEncoding attribute: I'd prefer a method where I'd pass in a Node and get the encoding that would be used. Don't like the statefulness of the attribute.
Resolution: The LS ET decided to remove this attribute completely since it doesn't really serve any valid purpose. The LS spec will not define an API for finding out what encoding would be used for a particular Node.
Issue LS-Issue-75:
DOMWriter.errorHandler: Might be more general than just errors, could be reporting progress or other details (such as the selected encoding) or participating in filtering.
Resolution: No, the error handler is an error handler and nothing more. Other API's should be defined for things like progress notifications or other such callbacks. Unless someone provides a compelling usecase for changing this, it won't change.
Issue LS-Issue-76:
DOMWriter.newLine: Should probably be a unsigned short with constants for the supported values like other enumerations in the spec.
Resolution: Description updated, this will remain a string and the definition was relaxed to support any string so that future unicode newlines n' such can be used w/o an API change.
Issue LS-Issue-77:
DOMWriter.setFeature method: Should have an defined exception for inconsistent features, like turning pretty-printing on after setting canonical-form to true.
Resolution: See issue 90.
Issue LS-Issue-78:
DOMWriter.writeNode method: Writing a Document or Entity node... well formed XML. Why would writing an entity node be well formed XML?
Resolution: Description updated.
Issue LS-Issue-79:
DOMWriter.writeToString method: How is this affected by encoding? It will be represented internally as UTF-16 on most binding, but users who have set encoding to ISO-8859-1 or US-ASCII might expect no code points higher than 255 or 127 respectively so they can naively write out the string to a file later.
Resolution: writeToString() always writes into a DOMString, which means it's always UTF16. The encoding information available is always ignored in writeToString(). Description updated to reflect this.
Issue LS-Issue-80:
DOMInputSource Interface: I don't like the multiple personalities of this interface. Instead of creating a DOMInputSource and then customizing it by setting attributes, I'd prefer multiple create (createSourceFromURI, createSourceFromString, etc), methods on DOMImplementationLS and only the minimum read-only attributes on DOMInputSource.
Resolution: Won't change, there are too many combinations of input sources to define specific factory methods for all combinations.
Issue LS-Issue-81:
DOMEntityResolver Interface: "for applications that use URI types other than URIs" Did you mean URL's.
Resolution: Description updated.
Issue LS-Issue-82:
DOMBuilderFilter.acceptNode and .startContainer: If the return value was a Node, then a Filter could:
  1. return the passed enode to have the element inserted.
  2. return null to have the element rejected
  3. return a DocumentFragment for SKIP

Resolution: Won't change, this would make it more complicated and more expensive to implement than with the current proposal.
Issue LS-Issue-83:
DOMBuilderFilter.acceptNode and .startContainer: substitute a replacement element created with Document.createElement[NS]
Resolution: No, such mutations to the tree from a filter is not allowed by this spec.
Issue LS-Issue-84:
DOMBuilderFilter.acceptNode and .startContainer: It should be possible to throw an exception in acceptNode and startContainer to stop the parse.
Terminating parsing from a DOMBuilderFilter: The description of the DOMBuilderFilter states that parsing can be terminated early using a filter, but doesn't give a specific recommendation or mechanism regarding how to do this. Should this be binding-specific, or is there a particular DOM exception which should be raised?
Resolution: Use FILTER_INTERRUPT if you want to stop the processing of the document. Interrupting the processing of the document does no longer guarantee that the entire is XML well-formed.
Issue LS-Issue-85:
DocumentLS interface: An isLoaded or ReadyState attribute would be strongly desirable to determine that an async document was loaded without registering an event listener.
Resolution: This has been discussed and proposed before, and so far all proposals have been turned down. The load listener can be used for being notified about when a document is done loading, that lets you do everything a ReadyState or isLoaded attribute would do for you, cleaner and more efficiently (i.e. no polling of state, or anything like that).
Issue LS-Issue-86:
DocumentLS.load: Should an exception be raised if you attempt to start a second async load when one is already in progress?
Resolution: No, no exception. Calling .load() while a load is in progress on that same document will cancel the current load and start the new one.
Issue LS-Issue-87:
Document.loadXML: How would any XML declaration specifying an encoding be handled.
Resolution:
Issue LS-Issue-88:
DOMErrorHandler Interface: Called functions should be able to throw some type of exception or return an object to stop the parse and raise an exception to the caller of parse. Those exceptions would need to be added to the list of potential exceptions on the parse calls.
Resolution: Error handler methods can not throw exceptions. The main reason for this is that in the async loading case there's none on the receiving end of the call to the error handler that would be able to deal with the exception. And besides, exceptions are for exceptional cases, this would not be such a case.
Issue LS-Issue-89:
The description of the whatToShow attribute in DOM3 Load and Save for both DOMWriterFilter and DOMBuilderFilter is unclear. For example, if I set whatToShow to NodeFilter.SHOW_ELEMENT does this mean that only element nodes will be output? or does it mean that only element nodes will be passed to the filter for further consideration while other kinds of nodes will be output without being checked through the filter?
Resolution: The description is already pretty clear on this, no change.
Issue LS-Issue-91:
"entity-resolver": The description should describe what support a builder is expected to provide if the resolver is not specified. When a new builder is created, should a default resolver be exposed via this attribute, to allow client code to "wrap" a basic resolver, or should the default value be null? (This kind of information would be helpful for many attributes in the DOM spec.)
Resolution: by default the parser is free to do whatever he wants regarding entities resolution.
Issue LS-Issue-92:
DOMBuilder.errorHandler: When a new builder is created, should a error handler be exposed via this attribute, to allow client code to "wrap" a handler, or should the default value be null?
if no error handler, then throw exceptions?
Resolution: DOMBuilder.errorHandler might expose a default error handler at creation time. (changed DOMImplementationLS.createDOMBuilder description).
Issue LS-Issue-93:
DOMBuilderFilter.whatToShow: The description of this attribute states that attribute nodes will never be passed to the filter, and the description of the filter interface also states that the document element will not be passed to the filter. What about the Document, DocumentType, Notation, and Entity nodes?
Resolution: Document, DocumentType, Notation and entities are not passed to the filter.
Issue LS-Issue-94:
DocumentLS.saveXML: Why would the return value ever be null?
Resolution: "or null in case an error occurred."
Issue LS-Issue-95:
The DOMBuilder supports a "feature" called "create-entity-nodes"; is there a reason to also define "create-notation-nodes"? There's definitively less need to provide a filter of this sort. Perhaps there should be an option to not build the DocumentType node at all, even if present? "processing-instructions" ?
Resolution: Not enough use cases to include those features. Use DOMBuilderFilter. For DocumentType, implications on validation are not certain, so we don't plan to support it for the moment.
Issue LS-Issue-96:
The description of serializing character data and attributes is at variance with XML C14N rules; it seems preferable to stay consistent with C14N where possible, or at least to better motivate any departures.
For example, the description:
"Attributes containing quotes but no apostrophes are serialized in apostrophes (single quotes). Attributes containing both forms of quotes are serialized in quotes, with quotes within the value represented by the predefined entity "."
varies from C14N which never uses single quotes but always replaces a quotation mark in the attribute value with ".
Somebody should carefully review this text with respect to C14N rules, and either use C14N rules or provide feature options on DOMWriter that allows the user of DOMWriter to choose the appropriate serialization.
Resolution: use canonical-form for C14N, otherwise, we keep the way it is currently defined.
Issue LS-Issue-97:
Under the description of DOMWriter appears the following:
" When serializing a document the DOMWriter checks to see if the document element in the document is a DOM Level 1 element or a DOM Level 2 (or higher) element (this check is done by looking at the localName of the root element). If the root element is a DOM Level 1 element then the DOMWriter will issue an error if a DOM Level 2 (or higher) element is found while serializing. Likewise if the document element is a DOM Level 2 (or higher) element and the DOMWriter sees a DOM Level 1 element an error is issued. Mixing DOM Level 1 elements with DOM Level 2 (or higher) is not supported."
I'm not sure what this is saying. Is it describing a scenario where multiple implementations are simultaneously used with a single API and a document which was instantiated by a Level 1 implementation has an element which was instantiated by a Level 2 implementation? Wouldn't it be an error to import a Level 2 node into a Level 1 document in the first place? Or wouldn't such an import effectively downcast that Level 2 node to its Level 1 counterpart?
If, on the other hand, this language is not talking about multiple implementations, then how is it possible to have a Level 2 implementation create a Level 1 element? Any element created by a Level 2 implementation will be a Level 2 element.
Resolution: not an issue anymore (text removed).
Issue LS-Issue-98:
Regarding the "namespace-declarations" feature of DOMBuilder, which is defaulted as "true", meaning "include the namespace declaration attributes, specified or defaulted from the schema or the DTD, in the DOM document", how does this correlate with the following statements:
  1. in DOM-3 Core, under Element, it is stated "The properties [namespace attributes] and [in-scope namespaces] defined in [XML Information set] are not accessible from DOM Level 3 Core."; and
  2. in DOM-3 LS, under 2.1.3, it is stated "All information to be serialized should be available via the normal DOM API."

Unless I am missing something (which is probably the case), these latter two statements would seem to indicate that it is impossible to support "namespace-declarations" as presently defined.
Resolution: DOM 3 Core was fixed.
Issue LS-Issue-99:
DOMBuilder.parseWithContext: It states that the context node should be used for namespace resolution, does the same apply to default attributes and entity references, are these to be taken from the document on which the parse is done?
Resolution: default attributes and entity referecences are taken from the Document attached to the context node.
Issue LS-Issue-100:
Is document fragment going to be defined. Since you do not have to parse a complete document at that point, I suppose both
<foo/><bar/>
and
foobar
are valid fragments, but is there an exact definition for this? I am particularly interested whether a document type is allowed in the input source that is the argument of this method. Since the input may also be a document, I suppose the answer is 'yes', but I think that would require implementations (or maybe just mine?) to 'double parse' or at least examine the stream a little, as the fragments
 <?xml version="1.0"?>
 <!DOCTYPE foo>
 foo
and
 foobar
   

would have to be handled differently (one is wellformed xml, the other is not, and there is at least one parseWithContext-usage where an input with a doctype would lead to a wellformed result).
Resolution: parseWithContext can take an XML fragment: i.e. anything except an XML Document, a DOCTYPE, entities declarations, or notations declarations.
Issue LS-Issue-101:
reconsider your removal of the namespaces feature
Resolution: added back in the draft.
Issue LS-Issue-102:
DOMWriterFilter/DOMBuilderFilter: do you pass the document element, document type, document, etc. to the filter?
Resolution: Document, DocumentType, Notation and entities are not passed to the filter.
Issue LS-Issue-103:
Current proposal: parse raises DOMIOException. All DOMIOException must also be reported to the DOMErrorHandler.
do we need to report and throw the exception? report seems to be enough but if no error handler was set, doesn't seem right to die like that...
Resolution: No changes.
Issue LS-Issue-104:
the spec says DocumentLS "uses the default features". Does it mean that no validation, etc will be performed? basically XML 1.0 loading ? the document will be loading, attribute value normalization XML 1.0 will be performed, but no validation will occure (even if users previously set setNormalizationFeature() on the document)?
Resolution: entities - default is false, which is different from Core.
Issue LS-Issue-105:
DocumentLS.saveXML isn't clear enough as to whether it's deep or not. If the snode parameter is null, the whole document is serialized. OK, that's deep. But if it is non-null, then ONLY the Node provided is serialized? What if one wishes to use DocumentLS.saveXML to serialize a specific node AND its children? I think it might be nice to have a boolean deep argument for this method, and specify that the provided Node, and all its children, may be serialized if the deep parameter is true. If false, then only the provided Node is serialized.
Resolution: DocumentLS.saveXML saves the document or the given node and all its decendants to a string.
Issue LS-Issue-106:
There is no way to query the DOMImplementation for a specific DOMBuilder. support XML 1.1? support validation? namespaces=false?
Resolution: Issue moved to Core to let users enumerate all DOM implementations in the DOM implementation registry to find out if there's one that supports the specific features.
Issue LS-Issue-107:
Should we make DOMBuilder.filter, DOMWriter.filter, DOMWriter.encoding, DOMWriter.newLine a parameter on DOMConfiguration?
Resolution: No.