Explainer
What Is SGML, and Why Does Defense Documentation Still Speak It?
The Ancestor
Structure Becomes Law, 1986
SGML — the Standard Generalized Markup Language, ISO 8879 — is not so much a format as a constitution for formats. You define a document type: what elements exist, what may contain what, in what order. Every document is then validated against that grammar, and a document that violates it is rejected by the parser, not caught by a reviewer. Every idea this site keeps returning to — content models with teeth, validation as the floor of quality — entered industrial practice through SGML. HTML began as an SGML application; XML is its streamlined descendant.
Defense and aerospace adopted it first because they had the problem first: manuals the size of libraries, lifecycles measured in decades, and contractual delivery where “the file looks fine on my machine” is not an acceptance criterion. From the late 1980s, military standards mandated SGML-tagged technical manuals, and early issues of S1000D were SGML-based before the specification migrated to XML.
The Present
Frozen, Not Dead
XML, standardized in 1998, kept SGML's enforced structure and shed the flexibility that made SGML parsers a specialist trade. It won completely: current specification issues are XML-native, tooling is universal, and no new program starts in SGML. But sustainment doesn't care what's current — platforms fielded in the SGML era still fly, and their manuals, tagged to standards like MIL-STD-38784's lineage, are still on contract. That's why “SGML” still crosses our desks monthly, usually attached to the question what do we do with all of it.
The answer is the standard convert-or-ingest split — with one advantage worth naming: SGML libraries convert well. Because the structure was validated from birth, the mapping to a modern XML target like S1000D data modules is largely mechanical, unlike word-processor content where every document is a snowflake. Legacy SGML conversion is a core part of our technical publications practice, and the verdict we give owners of these libraries is the same every time: your predecessors chose structure early. Thirty years later, that choice is still paying you back.
FAQ
Questions We Hear
What is SGML, in one paragraph?
The Standard Generalized Markup Language — ISO 8879, standardized in 1986 — is a framework for defining markup languages: you write a document type definition (DTD) that declares what elements a document may contain and how they nest, and then every document is validated against that grammar. If that sounds like XML, it should: XML is a simplified subset of SGML, and HTML began life as an SGML application. SGML is where the whole idea of enforced, machine-validated document structure entered industrial practice.
Why did defense and aerospace adopt SGML so early?
Because they had the problem first: enormous technical manuals, decades-long lifecycles, contractual delivery between organizations, and zero tolerance for structural drift. Military documentation standards mandated SGML DTDs for technical manuals from the late 1980s onward — MIL-STD-38784 and the standards around it governed SGML-tagged deliverables — and early issues of specifications like S1000D were SGML-based before migrating to XML. When your manual outlives the software that wrote it, a vendor-neutral validated format is the only survivable choice.
What replaced SGML, and why?
XML, from 1998 onward. XML kept SGML's core idea — declared, validated structure — and discarded the flexibility that made SGML parsers notoriously hard to build (optional tag omission, elaborate minimization features). Being simpler, XML got universal tooling: every language parses it, every publishing system speaks it, and modern specification issues (S1000D's XML schemas, MIL-STD-40051's DTDs in their current form) are XML-native. SGML remains standardized but effectively frozen; nobody starts a new SGML program.
We have an SGML library. What are our options?
The classic convert-or-ingest split. Actively-maintained document families usually justify conversion to the current XML-based standard the program owes — SGML-to-XML conversion is well-trodden, and the structural discipline SGML enforced makes those libraries convert far more predictably than word-processor content. The settled long tail can be published as-is through a modern delivery platform while it awaits, or skips, conversion. What an SGML library should never be is a reason for shame: it means your predecessors chose structure early, and that choice is still paying you back.
Get In Touch
An SGML Library and a Deadline?
We convert SGML libraries to S1000D and modern XML standards, and we publish the ones that don't need converting. Sample first, quote after — never the reverse.