pat@pyuxqq.UUCP (Pat M. Iurilli) (04/04/84)
Why not give rmail the capability of redefining its destination?
This would provide a mechanism for changing a poorly routed reply to a much
better one. For instance i just got a message that was routed thru 16 systems,
and I could have eliminated 14 of them. If more people did thiis, wouldn't
this help ease unnecessary net traffic???
Pat Iurilli Bell Communications Research, Piscataway, NJ
{ihnp4,harpo,allegra,houxm}!pyuxqq!pathoney@down.UUCP (code 101) (04/04/84)
rmail re-routing is a good idea because it saves hops and probably time. rmail re-routing is a bad idea because it can lead to cycles, and thus infinite time. take your choice. peter honeyman
emigh@ecsvax.UUCP (04/05/84)
<>
Recently, I have noticed that the routing of mail has changed when
going through certain hosts. When I am trying to work out a path to
to a certain site, say ucbvax, I used to send mail to myself through
ucbvax as:
mcnc!decvax!ucbvax!decvax!mcnc!ecsvax!emigh
This allowed me to find the weak links in the various alternate paths.
In trying to reestablish a path to unm-la I tried this and found that
all paths through decvax short circuited the extra ucbvax!decvax part
and the effective path was:
mcnc!decvax!mcnc!ecsvax!emigh
As I recall, this may have happened with some other sites as well.
Although this makes my simplistic way of testing paths impossible, this
may be a better solution to your changes to rmail--have the forwarding
software look for the best path. Of course, for any one site, best
might be to reduce your long distance charges by routing it through
an unsuspecting site in your local dialing area!
--
Ted H. Emigh North Carolina State University Raleigh NC
USENET: {akgua decvax duke ihnp4 unc}!mcnc!ecsvax!emigh
ARPA: decvax!mcnc!ecsvax!emigh@BERKELEY.ARPArandy@utcsrgv.UUCP (Randall S. Becker) (04/05/84)
Pat,
Have you considered what the basis will be for rerouting a message?
Shortest path by the number of hops may work out to be a longer path
in terms of transit time. A simple and cheaper solution would be to
inform a user who sent mail via an inefficient route of that fact
and suggest a POSSIBLE set of alternates.
Randy Becker
USENET: ..!{allegra,ihnp4,akgua,...}!utcsrgv!randy
CSNET: randy@torontofair@dual.UUCP (Erik E. Fair) (04/06/84)
Wow, this is beginning to sound like (*GASP!*):
sendmail
Ask your nearest 4.2BSD guru for details. Void where prohibited (like at USG?).
Erik E. Fair
dual!fair@Berkeley.ARPA
{ihnp4,ucbvax,cbosgd,decwrl,amd70,fortune,zehntel}!dual!fair
Dual Systems Corporation, Berkeley, Californiaphil@amd70.UUCP (Phil Ngai) (04/11/84)
What if in addition to having rmail re-route, it also checked to
see if a msg had passed through its site before?
--
Phil Ngai (408) 988-7777 {ucbvax,decwrl,ihnp4,allegra,intelca}!amd70!philchris@basser.SUN (Chris Maltby) (04/12/84)
x What you all need is the wonderful Australian IMPLICITLY routed network. It works out the route for me. Of course, you would all need to get it to make any difference...
mark@cbosgd.UUCP (Mark Horton) (04/12/84)
Implicitly routed networks are usually nicer than UUCP style. However, every such network I have ever seen requires dedicated links, e.g. hardwired or leased lines. Can the Sydney network work with dialup lines?
fair@dual.UUCP (Erik E. Fair) (04/13/84)
The problem with the Australian SUN network is that it uses the wrong
address syntax. Now if they were sensible and had used an `@', instead of
a `:'... :-)
Erik E. Fair ucbvax!fair fair@ucb-arpa.ARPA
dual!fair@Berkeley.ARPA
{ihnp4,ucbvax,cbosgd,decwrl,amd70,fortune,zehntel}!dual!fair
Dual Systems Corporation, Berkeley, Californiaguy@rlgvax.UUCP (Guy Harris) (04/15/84)
Actually, it sounds more like Mark Horton's "uuroute"/"uubang"; "sendmail"
doesn't give the proverbial X about routing. It merely decides which mail
delivery program should be run to deliver the message; the delivery program
does the re-routing, if any.
Guy Harris
{seismo,ihnp4,allegra}!rlgvax!guy
P.S. We have a version that runs happily on a USG UNIX system (with UNET,
used for the SMTP daemon mode)bob@basser.SUN (Bob Kummerfeld) (04/20/84)
Yes. The Sydney Unix Network software can work with dialup lines. In fact it works with anything that can create a connection, give you a virtual circuit (but not necessarily a full 8 bit path) and close the connection at the end. The network is operating over fixed wires, leased lines, a semi-public packet switched network and Ethernet. As soon as we get our interface board we will have it running on a public X25 network. The system looks after the routing function, the user only specifies username and host as the address. We are currently adding domain handling. As well as electronic mail it also does file transfer from any machine to any other - multiple hops are no problem. The file and mail systems are just users of the basic network message transfer service. We have experimented with database enquiry services and plan to build a distributed directory service using the basic network. Higher level services are very easy to add. There are now over 85 systems around Australia with a new one appearing every week or so. UUCP is not the only way! Bob
bob@basser.SUN (Bob Kummerfeld) (04/20/84)
The actual address syntax is a function of the higher level protocol and has nothing to do with the basic network message transfer system. The network software sees the addresses split into user and host, it is up to the mail program to parse an address and give it to the network in the right form. I admit we made a mistake using ":" as the separator for our mail service. We did it because \@ is the kill character on our systems. Our mail programs now accept either ":" or "\@" now and we are tring to get people to use \@. We would plan to change over to ARPA syntax, domains and all. Bob.
lepreau@utah-cs.UUCP (05/07/84)
At the Salt Lake Usenix conference in the Wednesday morning session there will be a talk on the Australian network. The abstract sounded very interesting. Jay Lepreau