This code should now be EW SCNL compliant, utilizing tracebuf2 output.

watch  (watch.c, makewatch.*) was not updated to SCNL because of the existense
of sniffring and sniffwave, which are two working tools capable of providing
the same testing/verification functionality as watch.

DK 2005/04/20
##############################################################################


Before attempting to run this program for the first time you should:

1. Insure that earthworm.d has an entry for MOD_IMPORT_IDA

2. Insure that you have defined pin numbers for all expected streams
   (if you care about pin numbers).  See import_ida.d for words about
   the format of this file.

NOTES:

Three are three types of network I/O errors that the program considers
as transient (or at least potentially transient).  These are

1. timeout
   No traffic at all for 30 seconds (the server sends heartbeats every 15
   seconds, so no traffic means circuit problems, not simply no data
   available).

2. no connection
   This kind of error happens when there is a problem with your computer's
   connection to the network, or with your network's connection to the
   outside world.

3. can't resolve hostname
   Connections with the name server are slow, for example.

Connections that have been working OK but then run into one of the
above problems will normally be restored successfully in a few
minutes.  Connection failures due to any other reason are more serious
and may possibly never succeed.  The only instance I have seen of a
succesful reconnect for anything other than one of the above is when
the disk loop at the remote server is being reconfigured.  In that
case, the client can get a "server fault" error that goes away once the
work on the server is completed.

Network I/O errors are dealt with by import_ida as follows.  Timeouts
are dealt with transparently to the application by the idatap library.
The Xfer_Read() routine will retry on timeout until it succeeds or
until it fails for some reason other than time out.  If the reason is
(2) or (3) above, then import_ida will try at 1 minute intervals to
reconnect.  Failures due to any other error are also retried, but at 15
minute intervals.  This cycle will continue until successful for up to
one week.  After one week, the program gives up and closes its socket.
If you wish to restart a connection after a child has exited you should
first correct the source of the problem and then halt and restart Earthworm.

In order to avoid filling the log file with repeated entries noting the
connection failure each minute, only the first network I/O error is logged.
All subsequent failed attempts are done silently.  The eventual sucesss
or final failure is also logged.
