A. Pagaltzis wrote:
* David Powell <[EMAIL PROTECTED]> [2007-05-24 02:05]:
Wednesday, May 23, 2007, 9:11:09 PM, A. Pagaltzis wrote:
A client could then use this optional information, where it
is present, in order to optimise its processing.
RFC 4287 itself makes no such provision, though.
... and there is no guarantee that such an extension would
work, because by overriding RFC4287s assertion that a feed is
unordered, the extension would effectively have mustUnderstand
semantics, which Atom doesn't support.
mU doesn't begin to cover what's needed to support forward
compatibility. mU provides a means to crash already deployed code;
that's it.
How so? Pure RFC 4287 implementations would process the entire
feed as usual. Clients which understand the extension could skip
some of the processing based on the order of the entries. I see
no change in feed semantics that would affect oblivious clients.
Exactly. Not having to understand random markup is a *feature*. Atom
doesn't have the the uniform semantics needed to make as yet
non-existent "extensions" understandable. *By design* it's not Lisp;
it's not even RDF. [This is why I say xmlns provides modularity, not
extension.]
cheers
Bill
