Open chriskohlhoff opened 9 years ago
Just a comment to say the obvious alternative names are io_execution_context
and io_context
The name io_service
was chosen, after much discussion, during the Boost review of Asio:
http://lists.boost.org/Archives/boost/2005/12/98592.php
Prior to that it was called demuxer
. Other names considered include io_driver
, io_services
, io_broker
, io_system
and so on.
If there was to be a change, my vote would go to io_context
. I find it sits comfortably in design discussion, such as:
"All connections in the server run in the same
io_context
.""This program creates each socket in a unique
io_context
.""My server creates a single-threaded
io_context
for each CPU. Theio_context
for the next connection is determined in a round-robin manner.""An io_context provides a link between the networking API objects (such as sockets and timers) and the operating system facilities that implement them. By default, completion handlers for asynchronous operations execute in the
io_context
; this can be overridden on a per-operation basis, but theio_context
is the earliest point at which the event notification can be received."
@chriskohlhoff io_context is the new name in boost 1.66. Is io_service
deprecated and the typedef scheduled for removal?
Note to say that Detlef made the remark that he would actually prefer the term
service
to change. This may be worth exploring also.