Cheatsheet: XML

Well-formed means the document follows XML's basic syntax rules -- one root element, every tag properly closed or self-closed, attributes quoted, elements properly nested. Valid means the document is well-formed AND conforms to a schema (a DTD or XML Schema/XSD) that defines which elements/attributes are allowed and in what order. A document can be well-formed without being valid against any particular schema, but it can't be valid without first being well-formed.

For a DTD, reference it with a <!DOCTYPE> declaration and validate with a DTD-aware parser or a tool like xmllint --noout --valid file.xml. For an XML Schema (XSD), validate with xmllint --noout --schema schema.xsd file.xml, or the equivalent validation API in your language's XML library (e.g. Python's lxml.etree.XMLSchema, or Java's javax.xml.validation).

No -- a CDATA section ends at the first ]]> it finds, so you can't nest them and you can't include the literal sequence ]]> inside one either. If your content contains ]]>, split it across the boundary (e.g. write ]]]]><![CDATA[> ) or escape the affected characters with entities instead of using CDATA.

XML is case-sensitive and requires every element to be explicitly closed (either with a matching closing tag or a self-closing /> ), has no predefined tag vocabulary (you define your own elements), and must be well-formed or a conformant parser will reject it outright. HTML (as HTML5) is more lenient: tag names are case-insensitive, some tags can be left unclosed, and browsers use error-recovery rules instead of rejecting malformed markup. HTML also has a fixed, predefined set of tags with built-in semantics/rendering, unlike XML's custom elements.

Use XPath. //title selects every <title> element anywhere in the document, book/title selects <title> children of <book> only, and //book[@id='b1'] selects <book> elements with a matching id attribute. Most XML libraries (lxml, Python's ElementTree with a subset of XPath, Java's XPath API, xmllint --xpath) support these expressions directly.