Dare Obasanjo wrote:
> From my perspective as someone who now works on a blog service
> (http://spaces.msn.com for those that don't read my blog) I doubt
> that we'd want to put something like ['indicate this post has
> had significant changes in the Atom feed'] in our UI.
This would seem to imply that if spaces.msn.com supports
atom:updated, it would do so as though atom:updated were defined as
atom:modified is. In other words, *any* change to an item, no matter how
small, would result in a new atom:updated value. Is this correct?
I continue to be concerned that atom:updated seems to be an
overloading that will inevitably result in confusion. There are two bits of
information in any single instance of an atom:updated value. One is the
explicitly stated date of the update and the other is some kind of implicit
subjective and inconsistent inference hint concerning the importance of the
update. Since entry generators will not be using consistent rules for
determining "importance" and some, like spaces.msn.com won't even try to
have such rules, it is very hard to understand how a reader UI could do
anything useful with the "importance" component of atom:updated.
I know that I've said this before, but I think it would make much
more sense to remove the overloading by having the date of modification
specified by atom:modified and then a second optional mechanism for
explicitly stating the "importance" of the update. "Importance" could be
indicated by either an element or an attribute. i.e. one of:
<modified>...</modified><importance>high</importance>
Or
<modified importance="high">...</modified>
If "importance" is passed as explicit data, rather than implicitly,
then reader UIs can be written to handle those cases (rare or common, we
can't know yet) in which entries provide this value. There is, of course,
the issue of defining a common set of "importance" values. Defining anything
other than "low" or "high" would be, I think, an excellent task for an
extension effort. In any case, it is likely that the set of used values
would converge fairly rapidly (except for edge cases) once this feature is
let loose in the wild.
bob wyman
