|
|
On May 30, 11:03 am, Alexandre Ferrieux <alexandre.ferri...@xxxxxxxxx>
wrote:
> On May 30, 10:36 am, Stuart <bigdak...@xxxxxxx> wrote:
>
>
>
> > Perhaps somebody can help me understand this behavior..
>
> > On host1 inside the tcl shell I type
>
> > set s [socket -async host1 19000]
>
> > I get an error "couldn't open socket: connection refused". This I
> > expected
> > as I have no server listening on host1:19000
>
> > On host2 inside the tcl shell I type
>
> > set s [socket -async host1 19000]
>
> > and I get "sock5". There is still no server listening on host1:19000,
> > and so I'm
> > wondering why on host2 I don't get the "couldn't open socket:
> > connection refused" message?
>
> The reason is that the [socket -async] may fail synchronously on the
> local interface (pointing to the same machine), probably due to the OS
> optimizing the TCP handshake on the loopback. While when you're
> attempting the connection from host2, no such optimization takes
> place, and the connect() syscall returns success before getting the
> (failed) result from the handshake a finite amount of time later.
>
> -Alex
Thanks. That does make sense. When I removed the -async both the
attempts from
host1 and host2 failed in the same way with the connection refused
message.
Stuart
|
|