Dimou on mail configuartions

Date: Thu, 14 Feb 91 15:37:36 +0100 From: dimou@cernvax.cern.ch (maria dimou) To: timbl@nxoc01.cern.ch The following short note was prepared by M. Dimou on request by J. Richards and A. Silverman. It was read and commented by D. Wiegandt and it is sent to the mintlist for information. 24/1/91 --------------- Email policy for unix/ultrix workstations ----------------- - The owner or the manager of a newly installed workstation, which is supposed to be able to send and receive Email should update the sendmail configuration file with: . The complete domain information of the node. In ultrix systems for example this is represented by the value of the $D macro i.e. DDcern.ch . The name of the general purpose relay machine. Again in the example of ultrix systems this is represented by the $R macro and it should have the value 'dxmint', with the alternative of 'cernvax', as a backup relay, i.e. DRdxmint #DRcernvax . Arrange in ruleset zero (S0) for all foreign Email traffic to be routed via the 'tcplocal' mailer to the major relay machine (defined in $R in the case of ultrix and others) and for Email to local users to be delivered with the 'local' mailer. The reason for adopting this 'restricted' routing is to profit from the sophistication built in the sendmail configuration of the central gateway machines (dxmint and cernvax), which modifies bogus addresses, giving them a canonical RFC 822 format (i.e. user@host.domain). In case of a group of workstations, installed and operated as a 'cluster': . To avoid the risk of some of them being unavailable for incoming SMTP for extended time periods (for high incoming traffic a few hours are already a problem) it is advised to fake the hostname on the "From" address of outgoing messages, replacing the name of the individual workstation by the name of a well supported 'server' node. This is how Apollo's at CERN are organized, i.e. they are all addressed as 'user@cernapo.cern.ch' due to appropriate facilities in their sendmail. . For other types of workstations, where such a possibility doesnot exist, it is sufficient for the 'server' node to have the usernames of all 'client' workstations' users defined as aliases. . The other advantage that such a centralised management provides, is the availability of disk space for storing and forwarding the messages to the 'client' workstations. --------------- Email policy for VAX/VMS workstations ----------------- Unfortunately VMS mail is rather 'primitive', giving little possibility of protection against system management problems. As a minimum, owners of VAX/VMS workstations are advised to: . Publicise their Email address on a 'bigger' machine (which has a system manager watching on its availability and the users' disk space). . Set an autoforward on that node pointing to the VAXstation This autoforward should be cancelled when the station is shut down. M. Dimou CS/EN