On Tue, 6 Nov 2001, Matt Crawford wrote: > I've never been able to think of a theoretical reason why the two > endpoints of a point-to-point link have to have addresses which are > related in any way. Is there some more practical reason why they > should be?
The main reason I've found is that for cases where the link addresses are not learned dynamically via a routing protocol, having them related through a common subnet saves having to manually configure a static route. Consider an EBGP IPv4 example for illustration. Suppose you wanted to use unrelated addresses on a link between EBGP peers - for example they might be unnumbered links referencing and loopback or ethernet port address. You could do that but somehow you have to prime one routers RIB with the route to the other router so that the BGP session between them can start up - ie. a static route to the other routers link address needs to be manually configured. If however, you specify a common /30 subnet for each routers link address then the route between the two routers is implicitly created and no extra static route needs to be configured. Additionally the common /30 subnet allows an administrator to uniquely determine what the IP addresses of each end of the link is once one of them is known. This practice of using numbered links continues to be used for IPv6 EBGP links as well. While maybe not necessarry anymore due to builtin IPv6 neighbor discovery this practice continues I suspect mainly due to history and inertia. IMHO though, it's not a big deal. -------------------------------------------------------------------- IETF IPng Working Group Mailing List IPng Home Page: http://playground.sun.com/ipng FTP archive: ftp://playground.sun.com/pub/ipng Direct all administrative requests to [EMAIL PROTECTED] --------------------------------------------------------------------
