On 2 Nov 2004, at 19:36, Bob Wyman wrote:
Tim Bray wrote:It would be an incorrect simplification. An atom:origin is anWould it be an oversimplification to say that [PaceHeadInEntry] provides by value what atom:origin provides by reference? -Tim
atom:id, thus, it is not a "ref" it is an "id" and can't be reliably
dereferenced[1]. You can't use atom:origin to discover the contents of the
atom:head for the feed within which an atom:entry was published unless
you've either seen the feed before or are able to use some service that
would map from the id to a ref (URL) or would serve up just the feed head
data. (I am unaware of any such service and while we might build such a
thing, I don't think the utility of Atom should depend on it.)
Yes but what if we invented an atom:origin2 which allowed you to put the entire
contents of the head of the feed into the origin2 slot. Then you would have what Tim Bray was mentioning an atom:origin that would allow you to provide your header by value. Which pretty much is what you are proposing it seems.
Also, atom:id is an *optional* element of the feed's atom:head.
Also one could make it non optional. But that is not a problem because we are not
speaking about atom:origin, but about atom:origin2 where you can pass your arguments by value.
Thus, an aggregator may not always be able to construct an "atom:origin"
element for the feed without inventing a new atom:id (which would be a bad
thing.)
PaceHeadInEntry is particularly important for aggregate/synthetic
feed providers since a high proportion of the entries presented to the
reader of an aggregate feed will come from feeds that the reader has never
seen before. Even if a service existed to do id-to-ref or head lookups, it
would still be undesirable to use atom:origin in this application since
clients would end up generating a very large number of lookup calls to the
service.
Again there you are not replying to Tim's point. You are saying so and so is the
case with atom:origin as it is currently. Tim was asking if what you were proposing was not something like atom:origin2.
If it is something like atom:origin2 then atom:origin2 would fulfill your needs just as well, no?
bob wyman
[1] draft-ietf-atompub-format-03.txt at 3.6.1
