The Hubbub about PubSubHubbub
Wednesday, August 12, 2009 at 21:42 tagged live, pubsubhubbub, twitter, wordpress in Web 2.0, Web Development Ok, you have to love the name. Pubsubhubbub is simply the best name for a technology for ages, period. And it is currently pitted as the better step towards a Google Wave like future than Google Wave itself.
So what is PubSubHubbub:
A simple, open, server-to-server web-hook-based pubsub (publish/subscribe) protocol as an extension to Atom (and RSS).
In longer terms, it is part of a system that removes the need for Google Reader to query my RSS feed every few hours to see if there is something new, instead allowing for Google Reader to be notified when something changes. The argument is that it gives us the live web and it decreases resource need. For a better into check out This Week in Google Episode 2.
I really like it and so does Anil Dash as visible in his post The Web Way vs. The Wave Way.
There is just one thing that makes me wonder. I remember a long time ago when MovableType was it, with everyone arguing that it was better to have the blogging platform publish static HTML files when you published a blog post as you are publishing so few of them in relation to the traffic that the resources are better spent there. But servers got more powerful and Wordpress changed this around to say that you want easy and fast publishing, the possibility to change your design around, the switch things, which makes it better to have the publishing be dynamic. This is what I see again here. We had centralized systems and only a few of them but slowly we are getting more and we want to move around. This changes how the content should be put together. We have friends all over the place and hence on request we do not want them to come from one source but from several, which is starting to become possible due to server power and general infrastructure items like Pubsubhubbub. It's push due to the underlying changes in how we interact with each other and what we say is important.
We somehow want near-real-time communications. The problem I see though with Pubsubhubbub is how the technology scales if you have for example 1 million friends on twitter. With a centralized system this is relatively easy. If each of your posts needs to be published to 300.000 hubs, this starts to become a pain. But with a polling mechanism it would not be live. How much resources are we will to spend for live, and who will pay the bill?
Reader Comments (3)
Happy to hear you like the name! Good write-up of what your thinking about. I'm happy to see more in-depth posts like this. You do a good job of explaining the relevancy of these new protocols and "live" features.
There's some discussion of the "million fanout problem" in these threads from the PubSubHubbub discussion group which may help:
http://groups.google.com/group/pubsubhubbub/msg/665c284bf94e1cd4
http://groups.google.com/group/pubsubhubbub/browse_thread/thread/e2d5cbf6ff374b93/69eded7d7f931adc
Thanks for the comment and the links Brett. I'll need to get this all sorted in my brain to respond meaningfully to the thread but hope I get to it. Probably should read the specs first though, because the second problem I see is that hubs should possibly be emm... multihomed one might say. E.g. if I publish my thoughtstream to a hub it would go to several so that subscribers would stay connected if one hub goes down. Kind of like Google pulling RSS Feeds from two locations and merging them back together.
As said, need to delve deeper into this but there is some other stuff on my plate as well. Possibly a good read on the beach in 2 weeks ;)
But it was getting my interest so much that I blogged again ;)
http://watercoolr.nuklei.com/ If you're into pubsubhubbub check this out.