>As far as solidity...  If the endpoints are using TCP (stream TCP/IP), I
>would expect reliable connections with problems only if the network
>traffic and/or collision rate is so poor that you time-out.  If they are
>using UDP (datagram TCP/IP), that is a problem, but, still, if they are
>they should have the retry logic built in.  (I mention UDP only as it is
>possible, though not probable - I do not know what ADSM uses, but I
>assume its TCP, not UDP.)

The LMCPD, through which *SM talks to the 3494, installs the following service
definition in /etc/services:

  lmcpd         3494/tcp        # IBM Automated Tape Library Daemon

so it's TCP (connection-oriented) rather than UDP (connectionless).

As Alexander says, if your LAN cannot reliably conduct even TCP
communications, your site has serious problems that need correction, rather
than considering an RS-232 circumvention of those problems.

   Richard Sims, BU