The 200 EMA is one of the most popular indicators of all time with Forex traders the world over, and for that reason alone is worth noting due to the psychological effect on the market place price can have when hovering around the 200 EMA….

Over the years, we have many forms of this. IMAP and NNTP come to mind. They work so well, nobody even thinks about them anymore. Efforts like Marimba’s Castanet have come and gone. Newer stuff like Mesh also have a Pub/Sub model at Internet scale. The lowest common denominator is TCP, which was designed for this type of thing from the start. Open an socket, make a request (or not) and read 1 byte or many. You define your own protocol ontop of a bidirectional stream.

Things that have made this more difficult over the years is open sockets where expensive (over 1000 or so), and http started as a single request/response/close model – hence the typical model of client polling at intervals. ePoll is interesting, I will have to learn more about how that works to make open sockets “cheaper”.

Moreover, pure push is a white lie. In todays world where clients are behind firewalls and NATs, pure push doesn’t work. The workaround for this has been the client maintains an open TCP connection to a server in the cloud (ala GotoMyPC and others) and can read replies as they come in. Even with this model, there is still reconnection and network timeout issues that force a client to reconnect and “request” again. So everything still starts with an open “request”. You can do pure-push to public IP listeners and probably IPV6 clients, but those are not typical cases yet.

I think the more interesting work here is not Push as such (that is well-known art), but things like: type systems and dynamic types, how anonymous types can work, client Projections (i.e. Select), type and service discovery, standards, and remoting client Expression Trees (e.g. LINQ) and remoting client code blocks (For example, a client should be able to serialize a Linq expression to an Event Service to do their own Search -the Twitter Track problem). How to sort Expression trees to allow efficient server side processing and stop a search as early as possible. How can developers get publisher type systems into there IDEs in an easy way for developing and compile type checks?

May also want to see some other work in this area that is very interesting:

Reactive Framework (Rx) with Erik Meijer – Reverse IEnumerable
http://channel9.msdn.com/shows/Going+Deep/E2E-Erik-Meijer-and-Wes-Dyer-Reactive-Framework-Rx-Under-the-Hood-2-of-2/#Page=3