Generic Metadata Mechanisms

This document is obsolete.

For the current specification, see: HTML Standard: Microdata

There have been some requests for introducing generic metadata mechanisms into HTML5.

To help determine what we would need to add, and whether it is worth adding anything, we have to come to an understanding of what the goals and requirements are of such a proposal.

Please document arguments with links to supporting research or links to other wiki pages detailing the anecdotal evidence for or against particular aspects of the goals and requirements.

Goals

What is the problem we are trying to solve?

See Microdata Problem Descriptions.

Who faces this problem?

Currently a few groups. In the future metadata may become necessary for the average "web consumer" (human or machine) to sort actual information from presentation and structural cruft. In other words, a useful tool for determining the meaning, terms of use, quality and/or authority of a piece of data inside (X)HTML.

Requirements: If we assume that we are going to address this need, what do we need to provide?

Please demonstrate the reasoning behind each requirement, along with examples of how the requirements could be addressed.

Machine-readable

A machine-readable and standardized way to apply semantic properties (metadata) to DOM elements in HTML5 and XHTML5.

Disambiguation

These properties are capable of being disambiguated between multiple definitions of the property name.

Finding or defining meaning

We should be able to find or define an "authoritative" meaning for an abstract concept like "title" (eg. book title, job title, person's title, land deed, etc...).

Machine-usability

The metadata could be read by UA's and other tools to perform actions that would not be possible without "knowing" what type of thing, quantity, unit or quality an element represents.

DOM consistency

The DOM has to be consistent between the HTML and XHTML representations of the HTML5 specification. If it isn't, then migrating between the two becomes non-trivial, especially for scripting.

Ease of deployment

The syntax has to be something that Web authors can easily deploy. If authors can't deploy this, then it won't get critical mass and won't matter.

One could argue that tools will be used to deploy this, that it'll mostly be used by big sites like Facebook, and that thus individual authors don't matter, but this kind of argument ("the tools will save us") has been repeatedly shown to not work, because in practice the tools have to be hand-authored too, and so the complexity is just moved to other people.

Inlinability

It has to have a way to include it inline, so that it is quicker for non-professional developers to use and adopt. Also, putting metadata in the same location as content could prevent errors in updates or copying.

Abstractability

It has to have both a way to abstract it from the HTML, like JS or CSS.

Sustainability

Where possible the proposal should be resistant to temporary or permanent unavailability of an authoritative source (ie, vocabulary provider). This could be achieved, for example, through a P2P or DNS-like mechanism, or by not relying on external sources (e.g. in the way that SSL certificates are checked).

Not doing this would lead to failures during temporary outages or overloading of an authoritative source of metadata definitions, and may make it more resistant to hostile takeover or shutdown of authority.

Distributing an authoritative source needs not make it less authoritative.

Reuse

The proposal should allow metadata and authoritative sources to be reused across elements, pages and sites, because web developers are more likely to use something that does not require repetitively typing the same data.

Multilingual and Multicultural

Not all concepts can be expressed properly in English. A proposal should allow metadata for foreign languages and concepts.

Authority and Security

Since a potential use of metadata appears to be enabling future features of UAs and other tools it follows that this opens the end-user to additional risks. For example, could a page author or hijacker feed a virus to a tool by falsely claiming it to be another type of data; could harm be caused when a metadata authority is hijacked by a group to deliberately mislead or blackmail; could metadata be used for unintended purposes such as spying on or annoying users.

With these risks in mind should there be standard mechanisms for securing metadata and verifying its source (such as signing certificates, encryption or white/black lists).

Related Proposals, Research and Discussions

• WHATWG Discussions
• w3c Semantic Web Interest Group (SWIG)
• W3C SWIG Mailing List Archive
• GRRDL (Transformations of XHTML to RDF)
• RDFa vs. CRDF (Cascading RDF Proposal)
• Embedded RDF Wiki
• RDF in HTML (Embedded RDF Examples)
• Wikipedia page on Semantic Web
• What are Microformats? (microformats.org)
• Friend of a Friend Project (FOAF)
• Dublin Core Metadata Initiative (DCMI)

Pre-Existing Software Systems That Demonstrate A Need

• Operator - A semantic web processor for extracting metadata from all forms of HTML embedded by using Microformats and RDFa.
• Fuzzbot - A semantic web processor for extracting triples from HTML4 and XHTML1.0, 1.1 and 2.0 data sources.
• Longwell - Longwell is a web-based RDF-powered highly-configurable faceted browser.
• Piggy Bank - Piggy Bank is a Firefox extension that turns your browser into a mashup platform, by allowing you to extract data from different web sites and mix them together. Piggy Bank also allows you to store this extracted information locally for you to search later and to exchange at need the collected information with others.
• Solvent - Solvent is a Firefox extension that helps you write screen scrapers for Piggy Bank.
• Semantic Bank - Semantic Bank is the server companion of Piggy Bank that lets you persist, share and publish data collected by individuals, groups or communities.
• Crowbar - Crowbar is a web scraping environment based on the use of a server-side headless mozilla-based browser. Its purpose is to allow running javascript scrapers against a DOM to automate web sites scraping but avoiding all the syntax normalization issues.
• Referee - Referee is a program that reads your web server logs and crawls your referrers (the links that point to your pages) and extract metadata from those pages and text around the links that pointed to your pages.

Proposals

Inline (as multiple attributes)

Multiple new metadata attributes such as in RDFa.

• Pro: Reasonably simple to add to spec.

• Con: Dependent on changes to HTML spec for future changes to metadata spec.
• Con: Would probably require a different syntax for block or external version of same metadata (makes it hard to move).
• Con: Requires documentation and standardization in the HTML spec rather than through a separate document and standards body.
• Con: More potential for attribute name collisions with future HTML attributes.
• Con: Appears to make metadata reuse difficult.

Inline (in a single attribute)

One metadata attribute with complex content (such as the style attribute)

• Pro: New properties can be added without changing the HTML spec.
• Pro: Changing properties does not affect the DOM.
• Pro: The properties are grouped together.
• Pro: Requirements very similar to style="" and onclick="".

• Con: Requires new metadata format to be created.
• Con: Makes it harder to select individual property/value pairs through CSS or DOM scripting. (Might require dedicated APIs... Ugh.)

Block or external metadata

Metadata is defined elsewhere from element and targeted in the manner of CSS or Javascript.

• Pro: Does not clutter the HTML.
• Pro: Gives it more space to develop such as style did once it was abstracted from HTML.
• Pro: May be easier to import, export, reuse, sign and translate.
• Pro: May be applied to elements that the author cannot change attributes on (eg, dynamic, protected or generated content).
• Pro: Speed up where external metadata can be cached.

• Con: Requires new metadata format to be created.
• Con: CSS-like targeting or use of class or id to apply metadata adds complexity/indirection.
• Con: Extra HTTP-request