MRC@SU-SCORE.ARPA (Mark Crispin) (02/17/84)
It's a great example of why domains are a good idea and why using the return-path or other relative routes is a bad idea. Domains or any other absolute addressing system would have optimized the delivery route. You can't optimize relative routes, because there is absolutely no guarantee that the FOOBAR host at one part in the route equals the FOOBAR host on some other portion of the route. -------
steve@ucl-cs.arpa (Steve Kille) (02/17/84)
As there have been several comments on this, let me explain a little of what is going on. The critical system involved is COM, a conferencing system developed by Jacb Palme in Stockholm. There are two instantiations of this, which interact with the Arpanet: one at York, UK (yorkcom) which will send and receive using UCL as a relay; and one at QZ, Stockhom, Sweden (qzcom / oden.mailnet), which is a mailnet host using mit-multics as a relay. Both UCL and MIT use the '%' notation to hide non-arpa domains/hosts. There are some relevant points: 1) COM adds the _____(text123456) which Geoff noted. The number refers to the COM entry. 2) COM usernames have spaces. Palme found that simply quoting these caused a lot of problems with many Arpa sites (!!). Thus he chose to map spaces into underscores. Unfortuantely, the software is not perfect, and often lexical separators, as well as spaces embedded in usernames get mapped. This may be fixed at qz, but certainly not at york. 3) When generating reply addresses, the full trace information associated with the message is used to generate a humongous source route (which will get worse when a message is repeatedly answered). Attempts are then made to eliminate redundant components. (As can be seen from the phrasing of this, I believe it to be an inappropriate approach). 4) There is a bug/configuration problem at Yrokcom, which can cause cause strange messages to be sent to an address approximately relating to recipients in the message to: list. If this is not fixed soon, the York header-people entry will be removed from the UCL list expansion. Steve
Margulies@CISL-SERVICE-MULTICS.ARPA ("Benson I. Margulies") (02/17/84)
Mark, I agree. What I am still perplexed about is (1) why all that crap couldn't continue to stick around, quietly, in the envelope (and in a much more compact form) (2) if its in the envelope, why it is evil for not-very-clever hosts to use it if they are unwilling or unable to do better. --benson