Our network is at the start of a major change. In the next 2 years we are committed to changing to the international standard X25. As you probably realise, the protocols that we use to communicate over the network date back 8 years. At that stage there were no international standards and we were forced to create our own protocols: The carrier protocol NSI (Network Standard Interface), RJE (Remote Job Entry) and ITP (Interactive Terminal Protocol). Since then International standards have developed: X25 as a carrier protocol, FTP (File Transfer Protocol) and XXX (Interactive protocol- the symbols XXX stand for X3, X28 & X29 - the 3 relevent standards). Increasingly manufacturers are offering these standard protocols, and the network is now being expected to handle a variety of differing machines, as these new hosts connect we can no longer be expected to reprogram each of them to fit into our network. The first change that is being made is to bring our carrier protocol into line. We already have installed and working a small GEC X25 switch which we are using to test our own software. In the next month we expect to have working a PDP 11/60 as a Gateway between the X25 switch and our Cambridge Ring (and hence to the rest of the network). This gateway will look like the PSS Gateway but will not have the name/password accrediting of the PSS Gateway and so will be much more transparent to the user. During the next year we will be progressively shifting parts of the network, ie TCPs, Workstations and Front Ends onto the X25 switch in addition to connecting new hosts like the Vax to it. We will then reach a situation when only a few residual components will remain on the NSI network, the Cambridge Ring will however remain essentially the same. A change in the carrier protocol should not impact on the average user as it is simply a means of ensuring that packets are correctly delivered from one part of the network to another. However to achieve full integration with new xcomponents, eg the Vax with X25 & XXX software, we will need to either gateway from ITP to XXX or change the TCP's to XXX. This change will have an impact on existing users as there are two features of ITP that do not map successfully into XXX. The first feature is a prompt, in XXX there will be no way to suppress prompts in a TCP if a user has typed ahead. The second feature is an Interrupt, in a PAD (Packet Assembler/Disassembler - ie TCP) there is only the concept of 'break', which can be efficiently mapped to an 'Int:A' but means of handling other interrupts, both single and multi character can only be inefficient and very slow. On the other side of the arguement, type-ahead on other operating systems such as Vax/Vms is handled in a completely different way, VMS does not echo user input until it is actually ready to use it and this presents problems in mapping onto ITP that can be handled more efficiently by XXX. It is possible to change the Emas front ends so that they support both protocols, this however is impossible within a TCP, although it is feasable to have both ITP TCP's and XXX TCP's on the network. A decision will need to be taken in the longer term as to whether the attraction of XXX of easy access to a number of hosts overcomes the disadvantage of the deterioration in the appearence to Emas users. There is no similar case to support the RJE protocol, on the contrary there are very strong reasons for using FTP. The main advantage of using FTP between machines with file systems is that the transaction is fully handshaked between the two ends. If for example a user password is incorrect, or a user attempts to 'grab' a file which does not exist, then this information is passed right back to the other end in a standard format, in the case of Emas a Mail message is then generated for the user with the specific information. We have already had some experience of using FTP to communicate with Kent and the DEC 10 and the protocol is proving that it is much more reliable than RJE. The major problem is that an FTP implementation takes a lot more code and it fitting this within workstations will be difficult.