BLARSON%ECLD@USC-ECL.ARPA (Bob Larson) (02/28/86)
Changing the message ID is counter-productive in detecting mail looping. (Adding one if there isn't one isn't.) Take the trivial case where list A and list B have each other as members. List A changes the message ID, sends the message to B who changes the message ID and sends it back to A. A hasn't seen the message with B's new ID, so the process continues... This is also a very bad idea for mailing lists gated to usenet newsgroups. Usenet normally has loops purposefully put in its newsgroup distribution, (to reduce the time it takes to get any specific message, and to add reliability by redundancy) which are taken care of by unique message IDs. There has been a proposal (in testing, I think) for multiple arpanet to usenet gateways for arpa mailing lists, increasing this redundancy. This would be broken by changing message IDs. There are numerous good reasons for not expanding a mailing list on one host. One is the world isn't SMTP, so it won't discover the hard to find loops anyway. Another is some hosts probably will answer with local only addresses. (Do you know how to get mail to ECLD, KYLARA, and HARVEY? ECLA (usc-ecl.arpa) does.) Probably the biggest objection is the load it places on the originating host and the network. There are already problems with duplicated messages due to hosts crashing in the middle of sending a message to a long list. Bob Larson <BLarson@Usc-Ecl.Arpa> -------