On 18/10/2007, Stefan Tilkov <[EMAIL PROTECTED]> wrote: > On Oct 18, 2007, at 4:19 PM, Steve Jones wrote: > > Guessing at not powerful, RSS/Atom is a pull model, you often want > > Push as well in pub/sub. > > > > Steve > > I agree (unless one starts adding new verbs or re-uses an open HTTP > connection). More often than not, though, I would claim that a pull > model like the Atom/RSS/HTTP one may be a better choice in terms of > scalability, developer productivity, and possibly even real-world > throughput and performance, at least for a loosely-coupled environment.
Pull can work sometimes, but often doesn't scale as effectively as push. As an example the SMS model with its 9 million messages an hour (UK only) to a huge amount of individual devices would be a bit of a bugger if all 50 million phones in the UK polled every second to see if they had a text (hell even every minute would be a bit of a spike!). If you have a large subscriber community and either P2P comms (ala phone) or a filtering/targetting subscription which narrows down the set per message dramatically then push tends to scale better. If all published events are of interest to all subscribers then often pull works better. I'm not sure why pub/sub couldn't be loosely coupled in a push environment though. > > Stefan > > >
