XML (Extensible Markup Language) was developed by a working group of the World Wide Web Consortium (W3C) and published as a W3C Recommendation in February 1998. It was designed as a simplified subset of SGML (Standard Generalized Markup Language), the complex meta-language that also gave birth to HTML. While SGML's full specification runs to hundreds of pages with numerous optional features, XML's designers deliberately constrained the language to eliminate ambiguity and simplify parser implementation — the XML 1.0 specification is roughly 30 pages, and a conforming XML parser can be written in a few thousand lines of code.

XML's core design principle is self-describing data: every element has an explicit opening and closing tag, attributes are always quoted, and the document forms a strict tree structure. This strictness (called "well-formedness") contrasts with HTML's permissiveness, where browsers silently fix missing closing tags, unquoted attributes, and other errors. XML parsers are required to reject malformed documents entirely — there is no error recovery, which ensures that any successfully parsed XML document has an unambiguous structure.

Schema languages provide formal definitions of valid XML document structures. DTD (Document Type Definition), inherited from SGML, was the original schema mechanism but has significant limitations: it cannot constrain data types (a DTD cannot require an element to contain a valid date), uses a non-XML syntax, and has poor support for namespaces. XML Schema (XSD), published by W3C in 2001, addresses these limitations with a rich type system including built-in types (string, integer, date, boolean), derivation by restriction and extension, and full namespace support. RELAX NG, developed by James Clark and Makoto Murata, offers an alternative that many consider more elegant and easier to learn than XSD.

XML namespaces, defined in a separate W3C specification, solve the problem of name collisions when combining XML vocabularies. A document might contain both an HTML table element and a data table element — namespaces distinguish them using URI-based prefixes. While essential for interoperability, namespaces are one of XML's most confusing features, adding complexity to parsing and querying.

Two fundamentally different parsing models exist for XML. DOM (Document Object Model) parsing loads the entire XML document into memory as a tree of node objects, enabling random access and modification but consuming memory proportional to document size. SAX (Simple API for XML) parsing is event-driven: the parser fires callbacks (start element, end element, text content) as it reads through the document sequentially, using constant memory regardless of document size. StAX (Streaming API for XML) provides a pull-based alternative to SAX's push model. The choice between DOM and SAX/StAX depends on document size and access patterns: DOM for small documents needing random access, streaming for large documents or when only specific elements are needed.

XML's role in enterprise systems remains substantial despite JSON's dominance in web APIs. SOAP web services, XHTML, SVG, RSS/Atom feeds, Microsoft Office formats (OOXML), LibreOffice formats (ODF), Android layouts, Maven build files, and countless industry-specific standards (HL7 in healthcare, FIX in finance, XBRL in financial reporting) all use XML, ensuring its continued relevance for decades to come.