stacy@sobeco.com (Stacy L. Millions) (03/16/90)
This may be a silly question (sendmail is just starting to make sense to me), but is there a way to configure sendmail to look up users via the MB records in the nameserver? -stacy -- "Women. You can't live with them. Pass the beer nuts." uunet!sobeco!stacy - Norm Peterson _Cheers_ stacy@sobeco.com
lamy@cs.utoronto.ca (Jean-Francois Lamy) (03/16/90)
stacy@sobeco.com (Stacy L. Millions) writes: >This may be a silly question (sendmail is just starting >to make sense to me), but is there a way to configure >sendmail to look up users via the MB records in the >nameserver? In all likelyhood you'd have to add this yourself. And also decide what you want the MB records to stand for. I don't think there is much established precedent for using MB records, so you pretty much get to decide for yourself -- I don't think any commonly available mailer would go querying a remote site asking it for its MB records and acting upon the received information though conceivably things like that could be done (analogously to the way MX records are currently handled. Jean-Francois Lamy lamy@cs.utoronto.ca, uunet!cs.utoronto.ca!lamy Department of Computer Science, University of Toronto, Canada M5S 1A4
gmp@rayssd.ray.com (Gregory M. Paris) (03/17/90)
In an article stacy@sobeco.com (Stacy L. Millions) writes: > This may be a silly question (sendmail is just starting > to make sense to me), but is there a way to configure > sendmail to look up users via the MB records in the > nameserver? If you check the relevant RFC (1035 at this time), you'll see that MB, MR, MG, and MINFO are all "experimental" RR types. That said, coincidentally, I just happen to be experimenting with them. I modified 5.61 sendmail to look up MB and MR records. Coding-wise it's not really too tough. The real difficulty is in determining what's the "right" thing to do and in trying to anticipate problems (mail or alias loops, perhaps). For instance, the RFC says that the main use of MR records might be to provide forwarding for people who have left a domain. To me that implies that there might not be a corresponding MB record (it's in another domain, after all), yet if you check the 4.3 BSD BIND guide, it says there must be one. I coded it up so that the MB is optional, which seems preferable to me. Another question is how to interpret the NEWNAME value contained in an MR record. For instance, what if the NEWNAME is something like greg.paris.ssd.ray.com? Which dot do you convert to an @-sign? Perhaps this is the reason that the BIND document says an MB is required -- MB returning MADNAME, a hostname only, reducing the possibilities for confusion. (No, it doesn't eliminate them.) Another tricky thing is when to look up MR and MB records. Do you only look for them in the case of "user@domain," or do you also look for them when you have just "user"? The former might not buy enough to make implementation worthwhile, whereas the latter would raise the ire of users who wish to receive mail on more than one host. A possibility I was considering was to use multiple MB records. Say gmp wishes all mail to be stored on host "quahog" except for mail sent to "hilarious," which he wants stored there. The relevant MB records might look like this: gmp IN MB quahog gmp.hilarious IN MB hilarious The lookup procedure would be to query for MAILB info on user.thishost and honor that, if found. If not found, check for just user and honor that if found. Otherwise, use the traditional local delivery methods (check aliases, then check passwd, or no such user). I'm not sure if all this really buys that much, though. Is anybody interested in this, or is the reason why MB/MR/MG/MINFO are still experimental that it's not worth the bother??? Greg -- Greg Paris <gmp@quahog.ssd.ray.com> {decuac,necntc,uiucdcs,uunet}!rayssd!gmp Please call me Eddie if it will help you relax.
tim@destroyer.cc.umich.edu (Tim Howes) (03/19/90)
In article <1990Mar15.190444.14424@sobeco.com> stacy@sobeco.com (Stacy L. Millions) writes: >This may be a silly question (sendmail is just starting >to make sense to me), but is there a way to configure >sendmail to look up users via the MB records in the >nameserver? At U of M, we have added a couple of nameserver resolver routines to look up MB and MG records and return a comma separated list of addresses (which is what sendmail likes). Once you have these resolver routines, the changes to sendmail are minor (alias.c is modified a little). The new resolver routines are available along with some other nameserver stuff on terminator.cc.umich.edu in ~ftp/unix/bind4.8.1.tar.Z. For more information contact hostmaster@umich.edu. -- Tim
roberts@nimrod.wpd.sgi.com (roberts) (03/19/90)
> I modified 5.61 sendmail to look up MB and MR records. Coding-wise > it's not really too tough.... Lookup of BIND information (and lots more including Yellow Pages info) has been done in the IDA + BIND 1.5 extensions to sendmail 5.61. Here's the original posting as I saw it: > From Dan@dna.lth.se Fri Jan 26 02:10:26 1990 > Path: nimrod!sgi!decwrl!wuarchive!uunet!mcsun!sunic!lth.se!newsuser > From: Dan@dna.lth.se (Dan Oscarsson) > Newsgroups: comp.mail.sendmail > Subject: name server based sendmail > Message-ID: <1990Jan26.101026.5696@lth.se> > Date: 26 Jan 90 10:10:26 GMT > Sender: newsuser@lth.se (LTH network news server) > Organization: Computer Science, Lund Institute of Technology, Sweden > Lines: 20 > > I have now and then seen questions about use of name server from > sendmail.cf and if MB,MR.. could be used for alias lookup. > > For you who are interested to try, you can try my sendmail which is > available by ftp from lth.se in directory Network/Sendmail. > > My sendmail is: sendmail 5.61 with the ida patches > other small patches from the net > enhancements for name sever use and international support. > > The name server support concists of extensions to query bind from rules > in sendmail.cf (for example: test for an MX) and to use bind for alias > lookup instead of /etc/aliases. > > The international support concists of: use of iso 8859-1 in letters, > translation to 7-bits if the receiver cannot handle 8-bit iso 8859-1. > And special translation tables for special hosts (for example to > convert the characterset in a letter from a Mac into iso 8859-1). > > Dan I have built this version of sendmail, and it seems to work well. I don't think we need to re-invent the wheel. - Robert Stephens