-Nginx frontend with Python running for web templating<p>-URL requests are matched with Erlang patterns (and possibly some regexes with Nginx)<p>-Python/ Erlang bridge done with internal REST Api<p>-most unstructured data stored in CouchDB<p>-requests end up in RabbitMQ message queue<p>-how message queue and rules engine (collection of state machines) match up buyers and sellers is not explained<p>-how transactions between buyers and sellers are stored by appending to long files is not explained<p>-how either these long files are spread out to many machines, or how CouchDB is spread out is not explained<p>Interesting talk, but except for the buzzwords they didn't reveal much of their internals.
I wonder what Twitter would have already become if they had early adopted an architecture like this?<p>I'm not trolling. Erlang is clearly ideal for this kind of thing. I wonder what kind of advantage it will give them?