A version of the Tuple Protocol is in 0.8, but it's actually not usable - it has some bugs. The version in TRUNK is currently in use in production at my company.
The Tuple Protocol takes a lot of liberties with the on-wire format to save space. In can get away with this by using the compiled code's metadata as the descriptor for how the data is organized. There are a lot of little optimizations like storing a single bitvector of optional fields instead of field ids for each. The required fields don't have field ids or types at all. There are also no separators between fields at all. On Mon, Apr 9, 2012 at 3:48 PM, Piscium <[email protected]> wrote: > On 9 April 2012 23:18, Bryan Duxbury <[email protected]> wrote: > > > And you > > should check out the Tuple Protocol, which is something I created that > > strips out a lot of the backwards compatibility features for the purpose > of > > saving space in serialized structs. It is also faster as a side effect, > > though I don't know if it's 20% faster. > > Is this Tuple protocol in Thrift 0.8.0? I don't seem to be able to find it. > > Would you please say a few words more about this protocol, like how > are fields separated? >
