PBC/LAN Gateways
-----------------

     The current method of attaching multiple ports, which would normally be
for hosts, is to put an individual Data Interface Unit on the end of each 
link to the host (see fig n).  Even when this unit can be racked, as with
the Plessey exchange, this solution is hardly elegant.  Unless the exchange
is particularly flexible this leads to hard decisions being taken on the
maximum number of ports that can be allocated to a particular host along
with the corresponding setting up of the ??? groups.  The problem can be
partially alleviated by taking some of the connections via the interface
units on to a standard Packet Assembler Disassembler (PAD).  This PAD can
then be connected in the conventional manner to the local data network, be
it X25, Ethernet or Cambridge Ring.  In the Edinburgh case we have connected
a number of ???? on to a range of hosts and on to both ICL PADs (OSLUs) on
an Ethernet and JNT PADs on X25 (see fig n+1).  This 'solution' does lead to
problems with some users seeing a different user image and the problems of
overcrowding.  This particular solution has been done as an experiment to
test out the differing connections.

     A more elegant solution could be achieved either by constructing a
gateway within the exchange itself or on the end of a 2Mb trunk (?) link to
the exchange (fig n).  ICL are in the process of building such a device as a
valuable extra to the initial exchange and we had hoped that they would be able
to deliver the latter before the end of our project but that is not now going
to be possible.  The advantage of such a solution is that the data interface
units are now part of the gateway itself, problems with allocating ??? groups
to particular line speeds are removed and local overloading of a particular
group of lines is alleviated.

     There are still several questions to be decided such as where the echoing
of a ???????should be done, if it is done in the host then the connection still
appears to be the same as a direct connection but the echo will probably be too
slow for most users as it will need to cross the LAN in single packets.  IF
the echoing is handled at least most of the time within the gateway then our
experience is that the response is perfectly adequate but the image to the user
of the ????????? session is not identical to the directly connected terminal.
I have said in our experience as the set up is not logically different from the
one which we are currently running.

   The other solution, that of a completely integrated gateway is being pursued
by at least one other PBX manufacturer, the connection from the gateway is to
X25.  The same consideration as above will still apply and at the end of the day
it will come down to which fits better in the user's data network environment
and to the overall cost.