W3C home > Mailing lists > Public > www-tag@w3.org > February 2011

TAG work on registries (ACTION-511)

From: Larry Masinter <masinter@adobe.com>
Date: Tue, 1 Feb 2011 07:40:53 -0800
To: "www-tag@w3.org" <www-tag@w3.org>
CC: Philippe Le Hegaret <plh@w3.org>
Message-ID: <C68CB012D9182D408CED7B884F441D4D058EDDD653@nambxv01a.corp.adobe.com>
The topic of "registries" in general has been discussed on www-tag for quite a while. Many of the general issues, and possible solutions, are being discussed in specific for the MIME registries (media types, charsets), but the issues arise for other registered protocol elements used in the web (URI schemes, HTTP headers and error codes, vendor prefixes in CSS, ....) 

The general issues are a matter for all standards organizations. A "standard" is a frame of reference, whether it is the standard for a meter, the width of railroad tracks, the threads on screws or an Internet protocol. Groups get together and write definitions for languages, terms within those languages, protocols, and protocol elements.  Sometimes those definitions are long documents (those attempting to define 'HTML' or 'CSS' or 'XML') and sometimes those frameworks make reference to terms, where the terms have independent definition. In the case of independent definition, there usually is a 'registry' or some way that those who define a term within the larger language make others aware of the intended definition.

It was an early principle of web architecture to try to avoid using registries and registry processes as a way of defining terms, but to rely on the power of the web itself for "distributed extensibility". (You could say that before the web, the idea of hypertext was limited because the hypertext systems predating the web had a closed architecture for hypertext extensibility; this allowed referential integrity at the expense of an N-squared communication cost for web-content extensibility). 

In practice, though, there are situations where the overhead of using a full URI for extensibility (such as is done with XML namespaces) is deemed to be too cumbersome, and protocol designers prefer using registered shorter terms (or prefixes) instead. 

This sometimes devolves into the questions of "who runs the registry" and "what are the rules for allowing items into the registry, and thus what assumptions can one make about terms that are in the registry". Some of the elements of the debate are based on efficiency and how protocol implementations are developed and deployed, and perhaps others around the power relationships or lack thereof when it comes to defining registered values.  RFC 2434
http://xml.resource.org/public/rfc/html/rfc2434.html illustrates a few of the considerations.

My suggestion for how to move forward with TAG discussion of these issues is for us to simultaneously (a) maintain a list of issues around registries (which W3C protocols and open web applications require, have, or need help with registries), but (b) focus very specifically on the MIME registries, issues with them, and the proposed solutions in http://tools.ietf.org/html/draft-masinter-mime-web-info with the intent to generalize later.

(Sent to fulfill ACTION-511: Send email framing TAG work on registries).

Larry
Received on Tuesday, 1 February 2011 15:41:41 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 26 April 2012 12:48:30 GMT