🚀 go-pugleaf

In article <6241@hydra.gatech.EDU> rh26@prism.gatech.EDU
(Robert L. Howard) writes:
> This is probably a simple minded sendmail question, but here goes...

	<Query about mail setup on Suns - All suns forwarding mail
	<to a central "mailhost", which exports its ...spool/mail/
	< to the rest of the suns.

What we do is to put dummy aliases in YP for all valid mail targets
(i.e. login ids, aliases, and mailing lists) of the form:
	target: target@mailhost

Since the order of searching for a "local" id is:
	1) Check YP aliases
	2) Check Local aliases
	3) Check for .forward
	4) deliver to /whatever/spool/mail/user
the presense of the dummy alias in YP will cause all "local" mail on
the mail clients to be sent to the mailhost.

The sendmail on the mailhost must be configured to recognize its own name
and also all the names of it's mail-clients as "local names".  The mailhost
also has a real aliases file (lives in /etc/aliases) that contains the
real aliases, mail lists, what-have-you; not just the dummy aliases in
the YP map.

> How does sendmail (on machine A) determine if a particular user exists

On a mail client, it doesn't matter.  Since all valid mail targets are
given a dummy alias in YP map, it'll be passed to the mailhost.  On
the mailhost, I'd presume it checks the YP passwd map, assuming it doesn't
find something else in the local alias file.  It would also get the
home directory from there.

> The problem I imagine here is that
> home directories not on mailhost won't be searched for .forward.  (We
> do run the automounter.)

We don't run the automounter, so I'm not sure what that impact would be.

".forward" files are a problem area - I do not know exactly how sun
sendmail handles things, but this is certainly a tricky area.

	1) Sendmail typically runs as "root", and "root" gets "nobody"
	priviledges on NFS mounted file systems, so an in-appropriate
	protection mode on either a user's home directory or .forward
	file could prevent root/sendmail from seeing the ".forward" file.

	2) An NFS mounted file system may be temporarily un-available
	at the instant sendmail chooses to look for a .forward file;
	hence it may think that there is none.

Both these problems will result in mail being delivered locally to
spool/mail, when prehaps it should have been forwarded elsewhere.

I don't know how or if Sun sendmail addresses these problems.  I think
that by appropriate careful coding, sendmail could work around both
these problems.

-Larry Parmelee
parmelee@cs.cornell.edu