Home > Archive >  2009 >  July >  10

Google's Pubsubhubbub

Friday, July 10, 2009 by Dave Winer.

It's got a weird name, and I found the spec somewhat hard to understand. 

But thanks to Brad Fitzpatrick and Brett Slatkin from the team at Google that implemented it, I now understand what Pubsubhubbub does. 

It allows you to receive updates of RSS feeds without polling.  

It makes it possible to build a distributed Twitter-like system with components that are not made by a single company, and with servers not run by a single company.  

It makes instant updates possible for RSS. 

It makes it possible to build a Twitter without the limitations of Twitter. (For example, no 140-character limit, the ability to handle enclosures, categories without #hashtags.) 

The protocol it defines seems reasonable (I'll have to implement one side of it to be sure) and because it has the backing of Google, one of a very small number of companies with the resources to make something like this work, it has a chance of gaining traction and when it does, scaling. 

In fact, it's part of one of the components I asked Google to implement in a blog post here on May 28, as Brett pointed out in our phone conversation earlier today. It's nice to see that at least a few people at Google see the possibility of assembling a Twitter-like notification system with the Small Pieces, Loosely Joined approach.  

Drilling in one more level, here's how it flows. 

1. Any feed that wants to participate in this network must add a bit to the feed that indicates which ping server is handling notifications on its behalf. There can be more than one. 

2. When a subscribing application initially parses the feed and notices this bit, it sends a notification to each server saying "I want to be notified when this feed updates." 

3. When the feed updates, it pings each of the servers it has registered with saying "I have updated." 

4. The server then pings each of the subscribers saying "He updated." 

The subscriber must have a known address, therefore must not be behind a firewall or NAT. For client apps, they need some kind of proxy that has a known address. This limit is signficant, but certainly not insurmountable. 

I would like to see them understand RSS syntax in addition to Atom syntax, and I understand from the spec that that is forthcoming.  

Update: http://superfeedr.com/ has also implemented this protocol.