We restricted our attention only to things that can be identified unambiguously. Web documents, chat channels, netnews articles, and even MD5 message digests can all be named with URL-style identifiers. Our infrastructure will not support labeling of document subparts, individual contributions to chat channels, or other things for which there are no naming conventions.
The technical working group will define a format for labels, indicating required and optional fields. The format will not specify which words or categories will be used for labeling or the criteria for assigning labels to items.
We will define at least three label distribution methods:
Our agreements will address the issues of label authenticity and the contents of documents changing over time without names changing. Our solutions, however, may not be 100% foolproof on these dimensions. Where possible, we will rely on existing technologies and standards.
We specifically rejected consideration of label distribution through conventions for naming the document (e.g., put xxx in the URL to indicate material that contains nude pictures). These conventions are being pursued by others, and we are not arguing that they are bad. They are not consonant with our mission, however. First, such conventions would permit only one or a few labels per document, which privileges the values of those who get to choose the labels. Second, participants felt that such conventions would only be meaningful if they specified category names, which would not be values-neutral.
Our schemes will permit filtering either at an end-user's PC, or somewhere in the network. We will not specify user interface characteristics, though we may suggest to implementors functions that browsers should support.