Excerpts from Noel Chiappa on Sat, Apr 04, 2009 12:35:50PM -0400: > Note also (in light of some other comments here that I don't have time to > reply to in detail) that in general, the 'name of the location' is only > actually _used_ at the internetwork level (for getting user traffic to the > named location).
Clarification: it is only used _for location_ at the level at which it names a location and by functions which do forwarding. Consider MIPv6: from the point of view of a correspondent node, to start with a home address names, and only names, a point of attachment. Later on, when "route optimization" is done, the home address is also used as part of authentication and session identification. But the functions that use it in those ways are not part of IP forwarding. > It may seem that higher-level entities _also_ have locations; e.g. > if you have a application which is running in a stack has only a > single interface, that application also 'sort of' has a location. Bzzt. If you want to talk about "location" of an application, be very careful what your layers are, and your layer networks. You can talk about location of a web services server in a web services layer network. That is far above IP and IP is not involved. You can also say that to reach an application via IP you need to send IP packets to a particular IP-layer location. These two views are independent. > _However_, as soon as you say that a higher-level thing can be > reached through more than one place in the network (or 'attached to > more than one place in the network') - which is what multi-homing is > - then the thing no longer has 'a' location. Whether it has a location in its own application-level layer network is not affected by how many IP-level connections there might be. swb _______________________________________________ rrg mailing list [email protected] http://www.irtf.org/mailman/listinfo/rrg
