Re: Extension mechanism

[Ken MacLeod <ken@xxxxxxxxxxxxxxxx>]

I think making a distinction, outside of <content>, between what might
be "metadata" and what might be "data" is a mistake.

Virtually everything we're specifying already *is* metadata about a
resource.  From Danny's description of a <metadata> element for a feed
or an entry, that means we have metadata for our metadata.

I think the confusion comes from not knowing whether <entry> is the
resource itself vs. <entry> being metadata about another resource
(some entity logically or physically in <content>).

While it is true that RDF is very clear about this (for any given
subject URI, all its properties describe only that subject), RSS 1.0
actually has the same confusion we're talking about here.  RSS 1.0
says the subject (rdf:about) of an item "is a URI which identifies the
item", but it doesn't make clear whether "the item" is another
resource or whether the <rss:item> RDF description itself can be the
item.  Most common usage is the former, but the latter is often used
or proposed for things like tickers, playlists, or other ordered items
where the items themselves are the subjects -- many of the same things
proposed for Atom.

I believe the answer to the question of "is this metadata about a
resource or is this entry the resource" is centered around <content>.
If content is a complete entity (embedded or by reference), this entry
is metadata about the entity resource.  If content is not present,
empty, or is an entity fragment (the body of a log entry or comment,
for example), the entry itself is the resource (content is the body
value).

In _either_ case, direct children of <entry> describe _the resource_
(as just defined).

  -- Ken

P.S. +1 on a mustUnderstand facility.