Dave, even endless repetition of a false statement won't make it true...
Pito Salas of BlogBridge recently wrote that they have implemented "Reading List Pings." This brings to OPML the same push-based bandwidth saving technology pioneered and proven in the realm of RSS/Atom syndication. As Pito says: "As more and more people both publish and subscribe to OPML Reading Lists, the polling that we are all doing isn't going to scale."
Almost on queue, Dave Winer slams the idea of Reading List Pings while inferring that Pito and others are "doomsayers." Winer claims that he can produce a "mathematical proof" that polling does, in fact, scale. In the past, Dave has made similar statements repeatedly while arguing against push-based notification, but he is no more correct today than he's ever been. No valid proof can be concocted that shows that polling is more efficient than push-based "notifications." It simply can't be done. No amount of name-calling or accusing folk of being "doomsayers" can change the simple fact that polling has serious and unsolvable scaling problems.
Dave is certainly correct when he says that "HTTP has [more than one] very efficient mechanism for software to determine if a resource has changed." In fact, the most efficient of those mechanisms in use today is the one that I originally defined on this blog: RFC3229+feed. However, even the best polling methods do not result in a system that scales as well as a push-based "notification" system that only sends updates to subscribers if and when there are changes.
Dave claims that sending "a ping to every subscriber... isn't practical because of firewalls and NATs." He is simply wrong. For several years now, PubSub.com has operated a system that efficiently sends "pings" through firewalls and NATs with no problem. We rely on the JEP-0060 extensions to the open Jabber/XMPP Instant Messaging protocol to create cheap continuous connections from inside the firewall/NAT boundary. There are many other methods available for dealing with the firewall/Nat issues. For instance, users of KnowNow's products or any of the implementations of mod_pubsub accomplish much the same as PubSub (with less efficiency...:-)) by using persistent HTTP connections.
Dave claims that if you rely on a "central authority" (like PubSub...) "nothing is more efficient ... than eTags, nor as widely implemented, nor as utterly optimized." While Winer is certainly correct in saying that "eTag" based polling systems are widely implemented, he is completely wrong in saying that "nothing is more efficient" and that nothing is "as utterly optimized." The RFC3229+feed mechanism, which relies in part on eTags, is certainly just about as optimized as a feed polling method can be, but it is still vastly less efficient, less optimized and less scaleable than a well-designed push-based method. I know this since in addition to having defined the very efficient RFC3229+feed mechanism, I've also personally directed, at PubSub, the construction of not only one of the blogosphere's largest feed polling systems but also the blogosphere's largest push-based notification system.
Dave's lack of understanding of the issues related to scaling can be seen in the history of the weblogs.com site that he struggled to build and maintain for so long. That site takes "pings" from blogs and then consolidates them into tremendous "change lists" which must be polled. Essentially, this site converts an efficient push-based update notification system (pinging) into an inefficient polling based system. Weblogs.com, as Dave built it, didn't even support common methods like eTags or RFC3229+feed to improve polling efficiency and scaling. The result was that it simply didn't scale and was frequently incapable of providing the service levels that people expected. Only now that Verisign has taken over the site and dedicated much better engineering staff and much more hardware, has the weblogs.com service begun to be somewhat useful again. However, since weblogs.com is still based on the terribly inefficient polling of change-lists that Dave supports, it is still a far cry from being what it might be.
It is true that mechanisms exist in HTTP to make polling more efficient and more scalable, however, those mechanisms do not permit polling to be more efficient than push notifications. Repeating false statements doesn't make them true. If Dave continues to claim that he actually has a "mathematical proof" that polling can scale better than a push-based system, he should present it. If he doesn't demonstrate his proof, then his claims are no more credible than Bush's claims to "secret" evidence of WMD's in Iraq or Joseph McCarthy's claims to hold a list of communists in the State Department...
Dave, if you have a "mathematical proof," show me. Show us all!