Phil Howard <PHIL@UIUCVMD>
Tue, 25 Nov 86 16:52:23 CST
|
The following are ideas for the DISTRIBUTE protocols. This was once a
project Steve Goldsmith and I were working on called EXPRESS.
1. Each EXPRESS server will communicate with its adjacent servers by
means of card image format files of a reasonablt small size. These
files will be packets in the sense that they will not necessarily
coorespond to files currently being worked on by EXPRESS. A standard
protocol format will be established and made public so that similar
programs may be implemented on various systems.
2. A windowing method sequence control shall be used to acknowledge and
syncronize the data flow between servers. This will help recognize
lost packet files.
3. A password system will be used on each packet file to help ensure
that the data really comes from the adjacent server. An encryption
system may be used as an alternative for increased security.
4. Connection control will be end-to-end. This will be like TCP/IP except
that the time durations will be very long and will span over systems
and network outages, including intermediate and end server nodes where
possible.
5. When servers exist on the source and destination nodes, delivery of
files will be in the exact order submitted, from the perspective of
a single destination address.
6. Delivery of files to destinations on the same node as the delivering
server will identify the origin as if the server were the networking
server (for VM, EXPRESS instead of RSCS, but TAG data formatted same).
7. Acknowledgements will be optional.
8. Maximum packet file size will be negotiable between adjacent servers
dynamically. Packets between servers will not be the same as datagrams
between end-to-end connection servers. (ISO model)
I can go on and on. Any comments so far? Any flames?
|
|
|