khanna@ARGUS.STANFORD.EDU (Raman Khanna) (08/09/86)
For a while there have been speculations as to who is doing what. I thought that I will give a status of what Stanford hopes to accomplish by the end of Summer ( Sept. 20 ). Stanford ACIS is working on a version of MacIP that will support ftp. Starting with the Cornell's Aztec-C version of MacIP, we will merge SU-PC/IP implementation to this environment. Tom Malloy and Andy Maas are very hopeful of having it up and running by the end of Summer. Perhaps a fancy user interface may not be ready by that time. Telnet is already running at a couple sites on campus in an "alpha" test stage. Across the street at CSLI, Bill Croft provided this summary of what he is working on: (1) A new scheme for IP address address assignment and resolution. Instead of each user's copy of MacIP being "customized" with a startup file on his local disk, MacIP programs (Telnet, FTP, etc.) go through a dialog with the gateway (using NBP) at initialization time to find out the IP address of themselves and the gateway. Part of the configuration information for the gateway is a "range" of IP addresses (on the ethernet side) that the gateway has available for "temporary" assignment to client Macs. This address range does NOT depend on any subnet scheme, in fact the AppleTalk cable segment(s) no longer have to be assigned a unique IP net or subnet number. Once the gateway has "assigned" an address from its cache, it keeps it "alive" by periodically (once a minute) sending ICMP echo packets to the client Mac. When the client Mac exits MacIP (or crashes, or powers off), the gateway discovers this in a few minutes and can free that IP address for possible reassignment. In addition to the range of addresses managed "dynamically", the gateway could also support another (smaller) range of addresses which have "static" assignment to specific client Macs via the local customization file on the client. ARP (IP address resolution protocol), like the address assignment protocol mentioned above, utilizes NBP instead of the more limited single-segment-only style ARP that was used before. We expect to complete these mods in the same time frame as completion of the Cornell/Stanford/MacIP release. (2) We had some outlines for a centralized UNIX name binding and administration service, similar to the BBN code recently released. With the release of that code it now looks more likely that we will simply merge the new IP functionality into the BBN/Columbia framework. raman khanna