Schemas and Initiatives

After a few years of META tag free-for-alls, the librarians got involved.

If you've never talked to a full-time, professional, library-science- educated, certified geek-type librarian about metadata, you really should. They're experts in the stuff. Most of them wrote their master's theses on "issues" you've never even thought about. Metadata is their thing.

Librarians' biggest issue with online metadata is that it's all too chaotic and proprietary. Proprietary META is horribly inefficient, because it causes a search engine to miss information that isn't using that engine's tags, and forces web authors to learn a dozen different taggings schemes. Engine-specific tagging is one of the reasons so many second-tier search sites don't have good databases.

In Real Life, librarians solved this problem years ago, by all agreeing to use universal schemas like the Library of Congress catalog system or the Dewey Decimal System. You can learn one system, and understand how to find things in all the libararies that use that system. More importantly, publishers can label their books using a system, and have it properly indexed in libararies all over North America.

Since the mid-1990s, various groups of libary scientists have been trying to create metadata schemes for the Internet. Some of those metadata initiatives design "universal" schemas that could apply to any resource, while others create specialized ones (often based on the universal ones) that provide detailed metadata for specific subject areas. Both types of initiatives seek the same goal: Creating a "write once, read many" scheme that all web authors and web services can use. To date, no such initiative has gained widespread acceptance, although a few specialized ones have traction in their specialities.

While there are dozens (if not hundreds) of metadata schemas in developement or in use, only a few have recommended using the HTML META element to label documents. (Most metadata initiatives are moving towards RDF/XML.) Most of the initiatives that do use META tags have adopted the concept of prefixing: Each indexing term is prefixed with two or three letters indicating which metadata schema the tag belongs to. For example, here's a tag that's part of a mythical "ABC" scheme:

<meta name="ABC.keywords" content="sex, drugs, rock and roll">

This page only lists metadata initiatives that I've confirmed as having recommended HTML META tags.


A-Core is an apparently defunct initiative to create a metadata scheme that describes the other metadata in a document. (For example, an "" tag doesn't tell you who wrote a document. It tells you who wrote the document's meta tags.) The HTML version of A-Core has/had five elements, and used the prefix "AC".

Reference: The A-Core: Metadata about Content Metadata

Dublin Core Metadata Initiative

The best known metadata intiative on the Web, the Dublin Core began in 1995 at a workshop in Dublin, Ohio, and has been revised a few times since then. The current version of the DCMI element set has 16 base elements (which use a "DC" prefix when used in meta elements) and 30 qualified elements (which are more specific versions of the base tags, and which should begin with the "DCTERMS" prefix; see below for information about older versions of Dublin Core).

Dublin Core is noteworthy among metadata schemes in that it also includes a set of predefined "encoding schemes" that can be used in the scheme attribute of a meta element. Officially-recognized Dublin Core encoding schemes also use a "DCTERMS" prefix. For example, these two tags use the W3CDTF scheme, which indicates the dates are encoded using the same format used by the WorldWide Web Consortium:

<meta name="" content="2000-01-01" scheme="DCTERMS.W3CDTF">
<meta name="DCTERMS.created" content="2001-01-01" scheme="DCTERMS.W3CDTF">

(Earlier drafts of the Dublin Core standard encouraged concatenating qualifiers onto their unqualified equivilents. (For example, name="" instead of name="DCTERMS.created".) They also allowed page authors to create custom qualifiers, but the current versions of Dublin Core allow only the 46 official values, because custom values aren't easily converted to other formats like XML, and often don't "dumb down" (correlate with unqualified elements) well.)

If the content of a Dublin Core element is a URI, the element may be better expressed as a link element.

Reference: Expressing Dublin Core in HTML/XHTML meta and link elements

GEM Project

The Gateway to Educational Materials Project is a U.S. Department of Education initiative to locate and label online resources (such as lesson plans) for teachers. GEM uses several of the Dublin Core elements and eight of its own elements. GEM's elements use the prefix "GEM".

Reference: GEM Element Set and Profiles Documentation

Weblog Metadata Initiative

The WMDI is apparently an attempt to create a metadata scheme (using several Dublin Core elements, as well as new elements using the "WMDI" prefix) just for weblogs. The project appears to be dormant, and I'm not sure the proposed tags were ever used in the wild.

Reference: Specification for Encoding WMDI Data in HTML <meta> Tags Version 0.1, 10/20/02