ip@cns.umist.ac.uk (Ian Pallfreeman) (11/28/88)
Alan, | Does anyone else see this happen? There are two entries in the janet.chn file | that map uk.ac.man and uk.ac.manchester to umist.cn.pa; Piete Brooks says that | this does seem to reflect what's in the NRS - so is Manchester's NRS entry | wrong? Does any one have any suggestions? First let me assure you that the NRS is correct. The people here have *really* decided to use a Prime to serve the whole of Manchesters mail requirements. You'll notice that I included a "please insult your local mail guru" in the Prime's error messages when I spotted all the "reverse" domains coming in -- they were coming in from a number of places, including Cambridge, UCL and even Manchester itself. At first we suspected mmdf of being clever, i.e. mailing to uk.ac.man in the hope that it would recognise uk.ac.man.misspelling, but soon found sendmail doing similar things. A number of sendmail.cf's seem to have: R$*.computer-science.manchester.ac.uk $1ux.cs.man.ac.uk buried in there. I'm not too familiar with the TECO-with-a-headache style of sendmail.cf as yet, so not sure what this line does, but it looks a little odd to me. A colleague at Rutherford was investigating further, but, typically, he left, and since SERC can't be bothered to buy us an X25 I can't be bothered to sort out the problem for our Unix machines. On the other hand, I hear 1.6 is around -- anybody tell me where (Jem?). Cheers, Ian. {SERC/ECF Sys/Net Support, ian@uk.ac.umist}
sysnet@central1.lancaster.ac.uk (Network Manager) (11/28/88)
While getting UK-Senmail 1.4a going on our new Sequent Symmetry I've come across a funny in the way mail to uk.ac.man.* is handled. A BSD4.2 VAX running an older version of the C-Nrs program and UK-Sendmail 1.4a handles things like fred@uk.ac.man.cs.ux quite correctly; however our system translates this into fred%ux.cs.man.ac.uk%man.ac.uk@umist.cn.pa, which isn't exactly useful! The address man.psy is similarly mangled, and so I guess is anything starting uk.ac.man. Does anyone else see this happen? There are two entries in the janet.chn file that map uk.ac.man and uk.ac.manchester to umist.cn.pa; Piete Brooks says that this does seem to reflect what's in the NRS - so is Manchester's NRS entry wrong? Does any one have any suggestions? Options like using mmdf are not open to us - we have what the suppliers give us and can't change, so what we have has to work. We also don't want to have to apply manual patches to the chn and dom files every time we rebuild the tables. Alan Phillips Computer Centre Lancaster University (sysnet@uk.ac.lancs.cent1)
ip@cns.umist.ac.uk (Ian Pallfreeman) (11/28/88)
Jim, Thanks for the patches, but I think I've found the problem. Looking at the c-nrs stuff produced on the Prime, they've registered uk.ac.man with an AR of umist.cn.pa rather than just as an alternate name. I think this is the trouble. Can you confirm? Can I say, defensively, that this was sprung on me by UMRCC without even telling me they'd done it, so I assumed they'd got it right. Silly me... Time to go find my head-rolling gear. Apologies, Ian.
jac@doc.ic.ac.uk (Jim Crammond) (11/28/88)
There are two ways to fix this:
1. remove the entry for uk.ac.man from the table generated by c-nrs and
let the ftp system figure out the routing.
2. make a small change to chn/MAKE in the awk script for the janet channel
where it thinks that rules like
uk.ac.man.cs.ux uk.ac.man.cs.ux
can be taken care of by the "general rule" and thus doesn't generate a
sendmail rule for it. The problem is that a match for uk.ac.man is
found before the general rule is tried...
The change is simply to remove the two lines of code thus:
if ( $1 == $2 )
{ if ( $1 ~ /^uk\./ ) # general rule will catch this
break
printf "R$+@%s\t\t$@<$1@%s>%s.janet\n", $1, $1, $1
}
becomes:
if ( $1 == $2 )
printf "R$+@%s\t\t$@<$1@%s>%s.janet\n", $1, $1, $1
There is a catch with this: namely that (assuming your c-nrs program
generates an entry for every site in the nrs and not just the ones with
application relays) your sendmail.cf will become very large and may be
a bit slow. Probably not a problem with a Sequent Symmetry mind.
-Jim.
ip@cns.umist.ac.uk (Ian Pallfreeman) (11/28/88)
Fellow sendmailers, The NRS registration for uk.ac.manchester was indeed entered incorrectly, and will be corrected as soon as possible: it should have been simply an alternate name (forward only) but was registered with an application relay instead. Jim tells me this should solve the Manchester problem. (damn, I was looking forward to dropping that bomb on my way home :-) BTW, I'm told uk.ac.umrcc will also be re-registered to point at our poor little Prime later this week. I look forward to it with some anxiety. Regards, Ian. {SERC/ECF Sys/Net Support, ian@uk.ac.umist}
robert@computing-maths.cardiff.ac.uk (Robert Evans) (11/28/88)
There's a similar problem with sites at Bristol. For example, uk.ac.bris.cs is a valid Janet mail site but uk.ac.bris get mapped to bris.ac.uk@uk.ac.aucc too soon resulting in the address cs.bris.ac.uk%bris.ac.uk@uk.ac.aucc. As Jim says, the way round this is to remove the janet.chn entry that causes it. I then explicitly put the entry back by appending it to local.dom: # # bristol corrections $+@bris.ac.uk $2%uk.ac.bris@aucc.ac.uk $+@bristol.ac.uk $2%uk.ac.bristol@aucc.ac.uk A similar thing should work for Manchester.