there are two basic use cases from my understanding. First, the
optimization of the memory block allocated with one call. Next, the
effect of protecting the Server against malformed or garbage binary
packets, at least to some extent.
I agree those are good reasons to have framed transport.
Is there not currently a third use case of allowing large messages to be
sent in chunks -- no a "TChunkedTransport"?
I was originally led to this possibly wrong understanding because
cassandra has both:
# Frame size for thrift (maximum field length).
thrift_framed_transport_size_in_mb: 1500
# The max length of a thrift message, including all fields and
# internal thrift overhead.
thrift_max_message_length_in_mb: 1600
which seems to imply that a message can be made up of >1 frame. Maybe
that's more related to cassandra's way of using thrift?
jrf
