At 01:48 PM 5/16/00 -0700, David Schnepper wrote:

>
>What are the advantages of OOB over a separate socket?
>
>--- I have to speculate a bit on this, as the OOB implementation pre-dates
>my
>start with the product.
>- There are less sockets consumed -- and I think some platforms had a limit
>on active count of sockets.
>- I have this funky memory that under a classic server there was a separate
>program forked off that was involved with managing the separate socket
>(I may be getting my memory confused with gds_relay on this point).

I checked with Charley. Here's the scoop: The remote interface/server
always open a second socket to transmit the updated event block from
the server to the client. To get the attention of the client in the
case of non-threaded platforms the server sends an out of band message
(1 byte) to force a signal. Multi-threaded clients use a listening
thread that more or less ignores the out of band message.

Jim Starkey