>
> > It appears to work as expected, but I don't like specifying the domain
> > name explicitly all over the file as I know by experience that this
> > will cause inconsistencies sooner or later. I would definitely prefer
> > to use $local_domains or similar on the left-hand-side, but haven't
> > been able to figure out how to "match into" a colon separated list. An
> > alternative would be to to have a common source for the re-write rules
> > and local_domains, but I'm not quite sure how to set up this, either.
>
> Put the valid domain names in a file (or dbm file) and use a
> lsearch/dbm lookup to match them. This can also be used in the rewrite
> rule. Cleaning up can be completed by using a macro at the top of the
> file for local domains and just using that in the rest of the file.
Is it really not possible to search a colon separated list?
> > Sendmail optionally matches the "user" part of the address against the
> > GECOS field as well as the user id. To provide a similar
> > functionality, I have defined the following smartuser and custom
> > transport:
>
> Yuck. Supporting this sort of thing is a seriously bad idea. Make
> people get addresses right, otherwise you can add a person with a
> similar fullname and suddenly mail delivery gets rather indeterminate.
Unfortunately, our management don't agree with this. They want our official
e-mail addresses, i.e. the ones printed on business cards etc., to contain
full names, with the usual convention of replacing spaces with '.'. We
could of course use the same convention for login names, but that would be
a bit impractical. Actually, our current policy makes such user ids
illegal, so we may safely assume that there will be no conflicts here.
> The only thing I would do is have a process that builds an alias file
> from appropriate passwd entries.
I may be fuzzy, but this is exactly the kind of thing I'm trying to avoid.
I want to use Exim to make life simpler, remember.
>
> Fuzzy matching might be appropriate for generating error messages - ie
> "I don't know that user, here is a list of similar users" although that
> has privacy implications - but it has no place in a serious mail system
> for delivery.
The matching doesn't really have to be that fuzzy (unlike me), a direct
comparison between full names would do in most cases. We're reluctant to
bounce messages unless it is absolutely necessary, though, as it can easily
give people the wrong impression. We're in the computing and electronics
business, so we certainly don't want to look like we can't set up an
e-mail system properly.
>
> > Neither of the search types for system_aliases seem to match the
> > lookup employed by sendmail, which will always consult /etc/aliases or
> > similar, but include data from NIS if a NIS include directive (+:+) is
> > encountered.
>
> Add an additional director to do NIS lookups. This looks like another
> broken Sun thing - there is no need to emulate that.
Not really. I think this is the way NIS was designed to work. Information
form passwd and group is looked up in exactly the same way.
Using the alias map directly would not really be a problem, though, except
it looks like the "inst" format version I down-loaded from freeware.sgi.com
was built without NIS-support.
- If any of the people responsible for SGI's distribution reads this, I
would like to ask if there's any change for an updated version with some
more features enabled. I can quite easily compile the system myself, of
course, but other people are likely to need this, too. Also, I don't like
the idea of installing critical software like this directly from source. I
can create my own "inst" package as well, but that's starting too sound
like much work...
> > I also want to set up some hosts with a "null config", i.e. let them
> > to forward everything to a "mail HUB", but I am not quite sure how to
> > do this. I haven't studied the documentation closely in this respect,
> > though.
>
> Take the standard config you use, set
> local_domains =
What will happen if I do this on the "best MX" ?
>
> Remove local transports and all the directors.
> That should then just work. An alternative is to make a smarthost
> router to deliver everything to the hub so outgoing mail goes via it as
> well.
That probably won't be necessary. We just want a single host to receive all
mail to avoid certain NFS related issues.
> > The below lines are necessary to be able to do anything useful with
> > Netscape Messenger. Why?
>
> > receiver_unqualified_hosts = localhost
>
> You are sending to unqualified addresses - exim only allows that from
> the local (ie unix command line) injection normally.
>
> > sender_host_reject_relay_except = 127.0.0.1
>
> You are injecting via local loopback SMTP, so that needs to be allowed
> to relay.
I had pretty much figured that out already, but are there any good reasons
why these operations aren't allowed by default? In particular, are there
any risks associated with relaying from localhost?
--
- T. Lund
--
*** Exim information can be found at
http://www.exim.org/ ***
