Summary: For some reason Matt Mullenweg started paying attention to Dave Winer. It's one of the worst ideas he's ever had, and he's the idiot who wrote Wordpress, so that's really saying something.
After a year or so of uselessness and a few epic flameouts from Dave, they'll probably implement PubSubHubBub on Wordpress.com, and rightfully not waste effort trying to make it work for self-hosters. With any luck, they'll stop serving duplicate feeds and go pure Atom at the same time.
Wow no reason to be so angry -- it's just some pings going back and forth. There's no need to make it personal. Besides, I've had way worse ideas.
We've always planned to have a variety of update notifications available -- the first we did was Jabber/XMPP, now RSS Cloud, after this we'll do some others. I mentioned this in my post on the subject, which you can check out here:
We've supported other things before, like Atom Publishing, that to this day get almost no use (literally less than but we support them anyway because it can't hurt and might help. In general I would consider myself API- and format-agnostic.
The vitriol on my part was due to the Winerian aspect -- the man has a long history of fucking up everything he touches technically, and vindictively fucking over anyone less than sycophantic. Run away.
APP doesn't see much use in Wordpress because nobody uses desktop blogging clients anymore. However, it is widely implemented on both sides (even Microsoft!), something that will never be true for <cloud>.
'RSSCloud' is inelegant, idiotically-designed (he thinks traditional SOAP posted to a resourcey URL is REST!), doesn't help centralized aggregators scale, doesn't work with NATed clients, and worst of all it was specced/never-implemented/forgotten by Dave Winer 8 years ago. At least you were smart enough not to use rpc.rsscloud.org -- Winer can't be trusted to host anything for anyone (eg: weblogs.com).
Google, Bloglines, Facebook, Twitter, et. al. will never implement it. The only people <cloud> could possibly help would be the few running their own server-side aggregators. Implementing PubSubHubBub would actually help people -- doing this just feeds Winer's ego (and thus pumps up the drama).
Why the hell were you still serving duplicate RSS2 at all anyway? The only extant feedreaders that don't grok Atom were written by Winer (and he's the only user left!).
Last month we got 688,000 new blog posts via XML-RPC, so desktop clients are still meaningful. (XML-RPC API is also used by other folks like Flickr.) APP may be better, I don't have a strong opinion, but it's hardly used and in fact we had trouble finding clients to test with when writing support for it. Regardless, we try to have the best support for APP possible.
Same thing with feeds -- we offer every feed in a choice of 4 formats (two Atoms, two RSSes). It's the same content and there's no overhead to us (it's just a template file) so why not let people choose what they like best.
In a nut-shell, rssCloud's subscriptions are merely a way of redistributing pings to subscribers. We think this is still on the publishing side of the problem and does not simplify the life of a subscriber. With rssCloud, subscribers must re-fetch the feed to see if it's changed. In contrast, Hubbub delivers the actual changes to the subscriber so they have no more work to do. This makes it much easier to subscribe, and has some nice properties when it comes to scalability.
So in plain english without all the hype, it's RSS pushed from the server instead of pulled by the client.
How are they doing it over http without thousands of connections being open and closed constantly on the server per second? Are they just announcing to a central server and the client only polls the one server for any one of thousands of blogs they could be subscribed to?
The client has to have a public internet port available for XML-RPC or an HTTP post. The central cloud server sends pings to all of the subscribers.
It's a bad idea.
If you're going to do this, just use XMPP. Otherwise, use HTTP the way it was intended and just have the client poll the server. Why does everyone think this is such a bad idea? Polling works. It scales. It's simple.
I agree on XMPP, but there's a certain amount of pressure to keep this working entirely inside the web-stack as much of the blog-community (and especially the Wordpress community) relies on fairly limited shared LAMP hosting, where XMPP running in the background (or anything else, for that matter) just isn't available.
Of course, the cleaner solution is to make this XMPP "inside" for thos who can use that, and then make a XMPP-to-HTTP frontend for those who can't. But on the other hand, there's a good point in keeping things simple.
After a year or so of uselessness and a few epic flameouts from Dave, they'll probably implement PubSubHubBub on Wordpress.com, and rightfully not waste effort trying to make it work for self-hosters. With any luck, they'll stop serving duplicate feeds and go pure Atom at the same time.