On Oct 16, 2005, at 6:45 AM, Robert Sayre wrote:
Hmm, in fact it seems positively perverse to require clients to
generate atom:id values with Atom's stringent uniqueness constraints,
when servers are typically better qualified to do this.
It's easy for the client to generate one, and probably a good idea. My
client follows the algorithm in <http://www.jwz.org/doc/mid.html>, but
changes the format to a tag:// URI.
Hmm, would someone familiar with "dumb clients" e.g. cellphones speak
up on whether this is something they should reasonably be asked to
do? I suppose they could use their telephone number rather than FQDN.
In any case, I would suppose that there will be many publishing
systems that will have their own requirements for atom:id, and thus
the protocol will allow atom:id to be overwritten on the initial
POST. For example, server systems that actually really believe in
http://www.w3.org/TR/webarch will want to use the post's HTTP URI as
its atom:id. Ruling that out would be violently unacceptable.
Given that, it seems perverse to *require* clients to supply it. -Tim
