On Jul 10, 2007, at 13:54 , Tobias Lütke wrote:
Before finalizing the protocol I think it would be smart to either implement this facility and design the binary protocol with this in mind or formalize a future safe extension mechanism.
All the details aren't in a spec yet, but the binary protocol is extensible. There's a single byte for commands, and this is a natural place to provide extensions.
There were a couple of interesting multi-invalidation options briefly discussed. We should figure out which one makes the most sense in a different thread. Choosing how to express it within each protocol is the easy part. :)
-- Dustin Sallings
