An application may need to perform small irregular changes in an XML document. It can be time-consuming to create a modified copy of a whole document for making each small change. It can be also hard to express condition for making the change in terms of XPath in order to use XSLT or XQuery.
DOM operations are the only operations that change an existing XML document instead of creating a new modified document. It is very important to understand that one document can be shared between many XML entities so changes made via one entity may affect data accessible via other entities that share the same document. Every XML document has special flags to "lock" its content and thus prevent the content from being changed. E.g. XSLT and XQuery processors optimize access to the source document in hope that every access to any entity will produce the same result during the whole run of the processor. If an extension function will modify the source document in the middle then the result of the processing is probably corrupted and this error may be very hard to detect and debug. To prevent this sort of problems, the XSLT processor locks the source at start and unlock at end so any attempt to modify the source during the run will signal an error. The document can be locked in the following cases:
The document is the source of the XSLT processing that is in progress.
The document contains the context node that is passed to the XQuery processor at start.
The document is used as a external generic entity in some other document.
The document is created by the call of XPath function doc, document or document-literal during the processing of some other document and that document is still in use.
One additional issue of using DOM operations is that any DOM operation invalidates DTD and schema validation information that was associated with the document before the operation.
Schema validation status ("is validated" flag). E.g. if an schema based instance of XMLType is validated against its schema and thus it has set its "is validated" flag to 1 then the flag is to be reset to zero if the XML entity of this instance is modified by DOM operation, because there is no guarantee that the modified document is still valid against the declared schema; if the application will repeat the validation and it is successful then the flag will be set to 1 again.
DTD-based processing of ID attributes. A validating XML parser can build special dictionaries of ID attributes when it reads a document. These dictionaries are used by
idXPATH function and "pointer operator" of XQuery to find a node labeled by a given ID value without searching the whole document. Any change in document tree invalidates data in these dictionaries so any access to them will signal an error.