While you can not be able to speed up your network links or reduce network latencies, you can at least arrange your message processing so as to recognize these issues. For instance, Internet sites can want to have two separate outbound TCP/IP channels: one for intra-domain traffic and the other for inter-domain traffic. Suppose, for instance, that your domain is claremont.edu. Then you might set up a new PMDF_SMTP_CLAREMENT generic or execution queue (OpenVMS) or Job Controller queue (UNIX and NT). Next add the rewrite rule

.claremont.edu $U%$H$D@TCP-CLAREMONT

and the channel definition

tcp_claremont smtp single_sys nox_env_to mx queue PMDF_SMTP_CLAREMONT TCP-CLAREMONT

to your PMDF configuration. This will cause all mail to other hosts in the claremont.edu domain to go out through the tcp_claremont channel. Processing jobs will run in the PMDF_SMTP_CLAREMONT processing queue. Internet mail for other domains will continue to go out the tcp_local channel and use the MAIL$BATCH queue (OpenVMS) or DEFAULT Job Controller queue (UNIX and NT).

You can apply this concept to other networks and channels. Indeed, you might find it useful to use separate processing queues for each of your outbound channels. This way, when one network is particularly slow, its processing jobs do not hold up jobs from other channels which can be waiting for the next available processing slot in the processing queue.