lear@NET.BIO.NET (Eliot Lear) (12/08/88)
What about using Class E addresses to partition the rest of the address space by network. This would work as follows: First, to handle a class E address, take three bytes of network data store them, and refer to an additional 4 bytes of data whose meaning would follow the same rules as the first 4 bytes. That is - class A, class B,..., and Class E. This way, you can chain addresses. Given a Class E address, consider the first three network octets as a network identifier on a slightly more global scale - allow them to identify internets. Allow large network gateways to advertise those first three bytes, instead of all of the networks currently being advertised. For example, let the milnet bridges advertise some class E address instead of net 26 and all of the networks connected to net 26. If an address is not class E, assume it is meant for the local internet. [The term ``local internet'' is sort of hard to fathom.] This should also allow for a smooth upgrade path, as noone should be affected as long as the existing networks are still advertised and addresses aren't reused. Whatya think? Eliot -- Eliot Lear [lear@net.bio.net]