mo@SEISMO.CSS.GOV (Mike O'Dell) (06/28/86)
It is clear the dreaded ISOOSI virus rotted the brain tissue of the IEEE committee, but why not solve the problem simply... (1) ARP is basically part of the mechanism for encapsulating IP datagrams on a local network. (2) There is a LSAP (or whatever) assigned for the DOD-IP suite. So..... Just change the "official" IP encapsulation to start with a 16-bit TYPE field, followed by whatever is defined by that field. With that, you can get ARP, you can get IP, you can get experimental protocols, you can get on with your life in spite of the crufty IEEE spec. Problems?? It breaks things. Well, so does their change. And it is early in the upheaval cycle. The other alternative is to get an LSAP defined for an EXTENSIBLE structure, with a field, and an Official Assigner. That has far more geopolitical problems than getting IP implementers to use the "sanctioned" IP encapsulation, but is desirable for more global reasons. Anyway, this approach would solve our problem, and move its administration back into a universe with lower entropy and more reality. -Mike O'Dell (I guess this is really a "secret" OLD ETHERNET type disguised as DOD-IP. So be it. Someone has to be clever around here. It sure ain't the people designing standards.)
LYNCH@USC-ISIB.ARPA (Dan Lynch) (06/28/86)
Mike (and others), This nasty little issue of living in a world not defined by us is one that can be meaningfully addressed at the TCP/IP Vendors Workshop at the end of August. At one time we will have in one room over 100 different TCP/IP implementation teams. WE can discuss not just the technical issues of the best way to cope with our bretheren in the ISO world, but how to make changes that are not disastrously disruptive to the customers in the field. (The most recent problem has been with subnetting and how it gets incorporated smoothly in the field. With over 100 different vendors it is impossible to coordinate release dates for protocol upgrades.) Meanwhile, perhaps some readers of this list have some experience running ISO IP as well as DoD IP on the same Ethernet and can reveal how they have dealt with it thus far. Dan -------