This isn't good.
This means that anybody who would use ZMQ to handle TCP connections
wouldn't be able to properly handle basic TCP error conditions because
they don't fit into ZMQ's paradigm.
Good architecture should handle things easily and with elegance. It
shouldn't be so abstract that it would ignore important life situations.
Cheers,
Yuri
On 9/27/22 22:25, Jim Melton wrote:
They aren’t reported because they aren’t “errors” in ZMQ. Because the
connection can happen independently (asynchronously) from the call to
to connect, and because a socket may be connected to a variety of
endpoints, the TCP conditions you are concerned about are not exposed
to you.
You are asking about detecting a condition that does not exist in the
ZMQ architecture.
Now, depending on the type of socket and the options set, there might
be an error in trying to *send* to a socket when the other end isn’t
present. Or it might not. PUB sockets, for example, need never have a
client connected to allow sends to succeed.
--
Jim Melton
_______________________________________________
zeromq-dev mailing list
[email protected]
https://lists.zeromq.org/mailman/listinfo/zeromq-dev
