When Exchange 2000 Server machines send and receive routing state updates, they use TCP port 691 for intragroup updates and SMTP on port 25 for intergroup updates. Why couldn't Microsoft use port 25 for both intragroup and intergroup updates?
The existing design achieves two purposes. First, the design meets the requirement that Exchange 2000 not break any existing messaging systems. Second, it makes updating routing group members more efficient. Recall that the routing group master must flood the routing group with link-state changes by sending updates to all routing group members as soon as the routing group master detects a change. (Fortunately, these updates are small; they're called "floods" because of their similarity to Internet Group Routing Protocol—IGRP—routing state floods.) As soon as any server in the routing group detects a link-state change, the server will use port 691 to tell the routing group master, which will in turn flood the rest of the routing group with the new link-state data. That method of notification works well for servers in the same routing group because one design approach to grouping servers in a routing group is to gather servers with uninterrupted connectivity together in one group. However, between servers in different routing groups, you can't assume such connectivity. Therefore, Microsoft chose to use the X-LINKSTATE2 verb as part of an ordinary SMTP session to exchange routing data between hosts in different routing groups. In that case, the link-state update doesn't occur until an actual mail transfer between the hosts takes place.