* The IMC will not start if you use a FQDN in the "Forward all mail to Host:" field and the name is not resolvable. So if you are doing dial-up and relying on the ISP for DNS resolution, use the IP address of the mail host here.  We believe that this is finally fixed in Exchange 5.5.

We have had several discussions with other folks in the past about this DNS configuration. We all agree there is some inherent inelegance in the solution since it is designed to take advantage of a failure condition. There are a couple solutions to this (besides a dedicated connection), but neither are really in the mainstream yet.

1. The ISP should configure its public DNS to list itself as the primary MX record for your domain so that the failure condition is not encountered by systems out on the net. Then they should configure their private DNS as described above. This way mail will always get at least as far as your ISP before it queues up. The problem here is that most ISPs do not run dual DNS trees, nor are they willing to set this up for only you. If they have it and it works, go with it.

2. The current Bind is 8.21 released June 21 1999. Support for RFC 2136 - Dynamic Updates in the Domain Name System has been added. If your ISP runs BIND 8.1, then you should be able to play some tricks with the DNS. The ISP would be the primary MX normally. But when you dial up, issue an UPDATE command to the ISP DNS server to make you the primary MX, dequeue your mail, change the DNS entry back, and hang-up. For more information on DNS and Bind, a good starting point is http://www.dns.net/dnsrd.

Our thoughts are that the ISP would provide a script that executed upon your connection that would issue the DNS update using the dynamically assigned IP address.   It is possible you could script it from the client side too, but then you would be responsible for issuing the update to change the record back when you disconnected - and in a disconnected state, you are in no position to do that.