socketTimeoutInterval is NOT for connecting

[Dar Scott]

On Monday, October 28, 2002, at 06:19 PM, RCS wrote:

>   I use 'socketTimeoutInterval' quite often...but that is for reading 
> and
> writing to a socket that is already open. I am looking for a timeout 
> setting
> for 'connecting' to a socket. If I use 'open socket' to connect to a 
> device
> that is either offline (or invalid), I would like to set a time out of 
> 5 or
> 10 seconds (the device usually responds in 1) instead of the default 1
> minute. I can only get so many cups of coffee in a day you know...

If the target computer refuses the tcp connection attempt some Windows 
systems will force a delay of 9 seconds.  If the target computer does 
not respond some Windows systems will wait for 21 seconds.  On OS X, I 
have seen delays around 75 seconds if there is no target computer.  I'm 
not sure, but I think if the connection is rejected it responds 
immediately.

(I have seen some problems in the callback never showing up in 
connections to the same computer or in connections by a very slow 
computer with Windows 2000 or XP.  This has been fixed in MC 2.5A7.)

On Windows, you will (now) get either the callback or a socketError().  
On OS X you will get either a callback or something else.  The something 
else has been socketClosed(), but that may have changed to socketError().

The solution I would try (actually, I will try, so I'm interested in 
what others make work) would be to use a send in time at the same time 
as the open.  If any of the callbacks come in first, they cancel it.  If 
it comes in first, it cleans things up.  I assume a close socket will 
clean things up.

An alternate strategy is to make very little dependent on the wait.  In 
addition, perhaps a cancel button (if appropriate) can do the work of 
the send-in-time instead of the send-in-time.

Dar Scott
(A Rev user sneaking in and sitting in on the back row and speaking up 
more than he should)