Using xml:base between an editing client and a publishing system
See atom-syntax on Relative URI references in editing clients. See IRC chat #1, IRC chat #2.
-
Somehow the UA now knows what one xml:base is to apply to content constructs so that rendered output references resolve correctly. This could be:
-
User enters "site root" value the user-agent copies to xml:base.
-
Publishing system supplies (in an as yet undefined manner) the default value the user-agent should use for xml:base.
-
Given a "site root" or host name, between the publishing system and the user-agent, the xml:base for the specific sub-site can be determined, for example, from http://foo.com/, to http://foo.com/user/md/us/bob/blog (see chat #2)
-
A solution for the publishing system to supply the xml:base is to use the "optional" support for getting back a "template" or empty Atom instance when you GET an Atom representation that has never before been PUT.
Using xml:base in Atom feeds
-
When a feed contains multiple entries, each with a unique xml:base (as would happen from republished feeds), the client must resolve all relative URIs before rendering in HTML, as HTML does not support changing the base internally.