20-07-2003
----------
When halting a server - if a timeout occurs kill off the process and remove the
request (and reply) queues.


15-04-2003
----------
* Not so much a TODO as a 'Please Take Note' *
DomBridge and NetServer send the originators message_queue ID through the round
trip.  The ID that gets back to the origination domain is used directly in the
msgsnd -> client call.  This is potentially very bad as it could be exploited,
and also means that different IPC systems (like FIFOs) cannot be used where a
platform port does not support SysV IPC.
However, in this case the 'msgqueue ID' can be interpreted as some index or
other.... Hmm. So in hindsight, as long as the originating domain knows what to
do with the 'org_r_msgid' when the reply comes in, we can intermix domains with
different IPC builds.
The only restriction is that the originator reply queue is represented
numerically in the request.

02-01-2003 *** DONE
----------
Read ENV from local domain and set in the environment for booted
servers.
OR provide a GW function gw_getenv(  ) which mirrors getenv but accesses 
the gwconfig

** Done -> new api function gw_getenv() for reading 'env' items from the
           configuration.  This was prefered to reading from the env.


29-05-2002
----------
Stats info should include requests sent/received/failed/success for any
given domain, also number of active semaphore/shm contentions.


24-05-2002
----------
Get configure to detect versions of mandatory Perl modules.


21-03-2002  *** Partially Done ***
----------
configure.in needs to find readline.h


13-03-2002
----------
- spool command for administrator (ala sqlplus/mysql)
- allow msec resolution in log (if LOG_WITH_MSEC is specified)
- queue stats command ** Part Done **


03-03-2002
----------
BB should be locked at a higher resolution.
Need to introduce Hash table for domains, servers and services.

Servers need to clear their 'reply' queues (if they've made any calls
themselves) when they're about to wait for a request - there may be
stale replies in the queue if any timeouts have occured. 

IPC queue addresses should be stored in the SHM along with an 'in queue'
counter (semaphore protected).

Portal needs to check that the reply being passed through it is the
reply that the client is expecting.  If it is not, then the reply should
be discarded immediately and not allowed to consume Phoenix resources or
network resources.

If the client runs in NET mode, then timeouts come from the config file - 
along with other values found in SHM.
