I was a bit skeptical about a new data format call tnetstring. It claims to be both human and machine readable. They claim that a parser is easier to built for a tnetstring and therefore is better. In fact, while the implementation of the parser may be easier, human readability suffers with negligible speed gains.
I assert that if human readability is harmed in favor of speed, then you would be much better off using a binary solution such as Protobufs.
Time is in milliseconds: