Networks 1983-1993
==================

(A brief look at some aspirations and problems)

The face of networks has been changing over the past few years and will
change further with the advance of technology.
Networks can be broadly split into two main types - the very high
speed Local Area Networks which operate over short distances and the much
slower Wide Area Networks for long distances.
    There are various types of LANs available, the primary types being
Ethernet, the IBM token ring and the Cambridge ring.  There are also a
number of other LANs made by particular manufacturers for their own
equipment, these are only similar in their inability to interwork with any 
type.

    In the context of WANs the existing packet switches are dramatically
increasing in power and there are two major new entries in this area - the
PABXs; capable of switching digital signals as well as the
conventional voice and the Broadband 'bus' carried over video cable.
     Each type of carrier has its own specific problems, the Cambridge Ring
uses active taps (entry point to the LAN), it shows instability with a large
number of connections and will probably price itself out of the market.  Ethernet
uses passive taps that are easy to add but its cable is both expensive and is 
difficult to lay.  A PABX does not utilise its bandwidth very efficiently and
hence requires links of a far higher capacity than are actually required,
finally, Broadband is expensive and time consuming to install.


Standards

      The main International Standards activity has been in the area of X25,
ie up to level 3 in the ISO reference model.  Although X25  and in particular
level 2 is satisfactory for conveying data over  potentially noisy long
distance synchronous lines it is not satisfactory for use over LANs,
the consequence has been that each LAN  uses its own protocols.  The emphasis on
standardisation has now shifted to level 4 - the Transport layer. There
is a UK academic standard, the so called Yellow Book but this will have to
be scrapped in order to match the emerging draft international standard
based on ECMA 75.  The ECMA 75 standard is more applicable for use over
LANs and hence will play a part in pulling networks back together but
it is always the case that standards activity trails along behind the
emerging hardware.
    There is still a lot of effort required to sort out the problems of
addressing over the various networks, too many networks have sufficient
addressing capability fot their own needs but rarely enough to carry an
address accross the next network or to allow the setting up of a unique 
addressing scheme.


The Future

     If our network is allowed to expand without control then we will see
a large number of differing LANs installed, the installation of each justified
in purely local terms.  If the LANs are then simply joined to each other by
Gateways then we will require a large number of gateways (N x N ?) which are
all slightly different.
    The problems of addressing and routing Mail and Files from a LAN
accross several other LANs to a destination will be immense, one obvious 
solution would be to include a Name Server on each LAN, but that would both
be expensive and difficult to keep up to date.
    With some control over the number of different types of LAN that are
installed the problems of addressing and routing will still be difficult
unless there is a strong central 'glue' to hold the network together.  
Under these circumstances a PABX will not act as such, it is closer
as it is a pure circuit switch and will not have the necessary addressing
range to force unique addressing on the network.

      One solution, already suggested for here and implemented elsewhere, is
to wire up the entire University for Broadband, this is a rather expensive
solution but would be one that is capable of the sort of expansion that our
users will expect in the future.

One factor in the future is that one of the current major prolems,
that of the cost of connection to the network, will not be eased.  Although
the power of the network 'pipe' to a user will increase dramatically is seems
that the installed cost will also increase.
                                                    B. Gilmore
                                                    Jan 1983