RE: Accuracy and Precision in dates

["Bob Wyman" <bob@xxxxxxxx>]

Sam Ruby wrote:
> Contemplate, for a moment, realtime publishing. ...
> consider the constraints of a user polling once an hour 
> or so.  In such a system, knowing precisely in which point in a 
> thirty second process the date was assigned amounts to unnecessary 
> precision.
	Realtime publishing[1] is what you get from PubSub.com if you use
our XMPP based service[2] and you are writing to a blog that pings us. The
time delay between when you publish an entry and when we deliver it to
subscribers can be on the order of just a few seconds...
	We should not be using the old-style RSS polling model as the basis
for designing fundamental components of the Atom format or protocol. If
anything, we should be assuming that publishing will be instantaneous -- the
format should work well even if publishing is orders of magnitude faster
than anything we have today. Let's not design in the constraints of old
technology...

		bob wyman

[1] Or, "near realtime" in order to keep the "realtime" purists happy...
[2] http://pubsub.com/developers.php, Note: Soon, we'll be announcing other
news aggregators that provide support for XMPP delivery of Atom entries in
near realtime.