[comp.mail.sendmail] Overexpansion of aliases

msf@prandtl.nas.nasa.gov (Michael S. Fischbein) (09/01/88)

Using the supplied sendmail on our systems (SUN 3/1xx 3.4, SGI 3030's GL3.6)
if you send mail to a defined group that group name is expanded in the header
and every recipient sees the entire list.  How does one change that
behavior such that the default, automatic behavior is to print the alias
used.  I know I can put the alias in the Bcc: field and I should get the
behavior I want, but I would like to set it up so it is automatic.

If this is relatively simple (as I think it should be), just email me;
I'll summarize.  Thanks,
		mike


Michael Fischbein                 msf@prandtl.nas.nasa.gov
                                  ...!seismo!decuac!csmunix!icase!msf
These are my opinions and not necessarily official views of any
organization.

msf@prandtl.nas.nasa.gov (Michael S. Fischbein) (09/02/88)

I've received several replies, and the answer is:

Aliases defined in the system /usr/lib/aliases are not expanded in the header.
Aliases defined in the user's .mailrc are expanded in the header.

Now, my sendmail is somehow getting my yellow pages domain as a host.
Oh well, I should be able to track that down.  Special thanks to
Creon@orville, pdb@sei.cmu.edu, and bob@acornrc.

		mike

Michael Fischbein                 msf@prandtl.nas.nasa.gov
                                  ...!seismo!decuac!csmunix!icase!msf
These are my opinions and not necessarily official views of any
organization.

dewey@execu.UUCP (Dewey Henize) (09/03/88)

In article <945@amelia.nas.nasa.gov> msf@prandtl.nas.nasa.gov (Michael S. Fischbein) writes:
}
(stuff deleted)
}Now, my sendmail is somehow getting my yellow pages domain as a host.
}Oh well, I should be able to track that down.  Special thanks to
}Creon@orville, pdb@sei.cmu.edu, and bob@acornrc.
}
}		mike
}
}Michael Fischbein                 msf@prandtl.nas.nasa.gov

We were told at a Sun 4.0 seminar that the yellow pages domain as host
is a new 'feature'.  Good luck.

Dewey Henize

-- 
===============================================================================
|             execu!dewey  Dewey Henize   or    juniper!dhenize               |
|         Can you say standard disclaimer?  I knew you could.  Somehow...     |
=============================================================================== 

karl@triceratops.cis.ohio-state.edu (Karl Kleinpaste) (09/03/88)

dewey@execu.uucp writes:
   We were told at a Sun 4.0 seminar that the yellow pages domain as host
   is a new 'feature'.  Good luck.

Misfeature.  Bug.  Bad bug.

It might be nice if I could have my mail-addressing domain name and my
YP domain name be the same.  I can't.  My mail-addr domain is
cis.ohio-state.edu.  A YP domain name of that length breaks PC-NFS,
which is so braindead that it apparently believes in a hard limit of
14 characters for a YP domain name.  (What did you do, Sun?  Develop
PC-NFS on a bunch of SysV-filesystem machines somewhere?)  So we must
necessarily hack our YP domain name down to something palatable to
PC-NFS.  Currently we use `cis.osu.'

John Gilmore (I think it was John; apologies if I am mis-attributing)
commented some time back that he had tried to argue Sun out of tying
YP domains to mail-addr domains and was unsuccessful.  It's time for
Sun to think this one through again.

A week or two ago, there was a question in Sun-Spots from someone
wondering why Sun seems to have gained such a nasty reputation in the
last several months (or couple of years, depending on the people with
whom you speak).  I have a theory for it: Because Sun is BREAKING
things left and right in the interests of `features.'

These `features' are not well-thought-out.
These `features' are DEEPLY BUGGY.
These `features' do NOT add significant advantage to my operation.
These `features' DO add to the number of things I have to manage.

All this is especially true with YP and the hassles of coping with
nameservers and mailers.  The nameserver is where host resolution
belongs; sendmail.cf is where mail-addr domain definition belongs.

The advantage of YP, for us, ends right at the point where management
of passwd, group, and similar files end.  Domain nameservice is best
done by /usr/etc/in.named, directly, and not by adding yet another
level of interaction between my application that wants, e.g., a host
IP addr and whatever-it-is which actually gives it to me.  YP should
be designed so as to GET OUT OF MY WAY when I find it does not fit my
needs.  Sun doesn't provide me with this ability.

I am tired of all of Sun, HP, and Pyramid deciding that They Know
What's Best For Me.  A couple of other vendors might fit the category
as well, but those 3 are the group with which I end up fighting the
most.  Sun is by far the worst offender.

An officemate suggested that I add the following: It seems that Sun
has been leading the way in so many areas for so long that they have
started getting cocky and are making too many of our decisions for us.
`They think because they have always done it right in the past that
they will always do it right in the future.'

Food for thought for Sun and other vendors,
--Karl

clarke@acheron.UUCP (Ed Clarke) (09/04/88)

From article <21282@tut.cis.ohio-state.edu>, by karl@triceratops.cis.ohio-state.edu (Karl Kleinpaste):
> dewey@execu.uucp writes:
>    We were told at a Sun 4.0 seminar that the yellow pages domain as host
>    is a new 'feature'.  Good luck.
> Misfeature.  Bug.  Bad bug.

[lots of good comments deleted in the interest of saving space]

The IDA sendmail patch kit and sendmail 5.59 source are available ( or 
will be soon ).  I haven't installed these packages on a SUN yet, but
I did see specific support for some kind of YP hack.  If you don't like
what Sun did, throw it out and use the IDA/Berkeley software.  That's
what I plan to do ...

Ed Clarke
uunet!bywater!acheron!clarke

deke@galaxy.ee.rochester.edu (Deke Kassabian) (09/06/88)

In article <298@execu.UUCP> dewey@execu.UUCP (Dewey Henize) writes:
>
>We were told at a Sun 4.0 seminar that the yellow pages domain as host
>is a new 'feature'.  Good luck.
>
>Dewey Henize

Well, I don't know about 4.0, but in previous releases (like 3.4,5) this
happened if and only if the host running sendmail couldn't resolve its own
name in the yp map hosts.byname (is the problem host a slave-server by chance?).  
So I don't think that this was being forced on you, but rather that it was 
just a sort of strange "default".  Of course, its a default you probably don't
want and may be able to get rid of by handling yp in a different way.
Good Luck.

BTW, I agree 100% with Karl Kleinpaiste regarding Sun's introduction of bogus
"features" to UNIX and networking software.  What we really want is support
for in.named without a layer of YP, and the option to have ftp telnet and other
software needing to resolve addresses to do so either by use of a host table
or by asking a nameserver.  It always seemed to me that it would be best to
have both... upon failure to resolve an address one way, try the other before
failing.  Just a thought.


  ^Deke Kassabian,   deke@ee.rochester.edu or  rochester!ur-valhalla!deke
   Univ of Rochester, Dept of EE, Rochester, NY 14627     (+1 716-275-3106)

nowicki%rose@Sun.COM (Bill Nowicki) (09/09/88)

In article <21282@tut.cis.ohio-state.edu>, karl@triceratops.cis.ohio-state.edu (Karl Kleinpaste) writes:
> dewey@execu.uucp writes:
>    We were told at a Sun 4.0 seminar that the yellow pages domain as host
>    is a new 'feature'.  Good luck.
> 
> Misfeature.  Bug.  Bad bug.

Let me try to clarify this.  In SunOS 4.0 sendmail, the $m macro is
pre-defined to be the YP domain name (with a mapping described in the
manual).  Thus IF YOU WISH you can have the exact same sendmail.cf file
on all machines, regardless of your domain.  This was necessary for our
own network, with five thousand machines in many sub-domains.  It was
just not practical to manually edit all five thousand configuration
files.  If you chose to not use this pre-defined macro, then you are
free to hack the sendmail.cf as before.  Old sendmail.cf files should
still work.  4.0 also had many of the IDA features like arbitrary
number of tokens in class matches, for domains of any number of levels.

> It might be nice if I could have my mail-addressing domain name and my
> YP domain name be the same.  I can't.  

One solution would be to create a short synonym for the long name.
Just put a single symbolic link in the /var/yp directory and the YP
server will support both domains.  This also aids in domain name
transitions.

> What did you do, Sun?  Develop
> PC-NFS on a bunch of SysV-filesystem machines somewhere?

No, it was developed on and for an archaic operating system called MS-DOS!

> I have a theory for it: Because Sun is BREAKING
> things left and right in the interests of `features.'

If there are actual things "broken", then please file a big report with
your service representative.  Emotional flaming on Usenet will not be
as likely to get your problem fixed, because we only have time to read
Usenet about once a month (too busy fixing all the reported bugs).

The original problem that started this discussion had nothing to do
sendmail.cf.  It was purely an aliases issue.  SunOS 3.x had a
"feature" that tacked on the YP domain name if a YP alias resolved to a
name with no domain part in it.  Since people complained about this
"feature" it was taken out in SunOS 4.0.  Upgrading to the latest
release would have solved this person's problem.  So would putting
domain parts into the alias file, (e.g. user@host instead of just user)
which is probably a good idea anyway.

> `They think because they have always done it right in the past that
> they will always do it right in the future.'

Not true.  But if you tell us what you want, then all we can do is fix
it in the next release.  For example, people have asked for direct
resolver support, so we will try to provide that in the next release.
Sorry, but we do not have a time machine to go back and fix releases
that were done years ago. Note that SunOS 4.0 DID have a version of
sendmail linked directly with the resolver. A shared libc with the
resolver was done by Hedrick@Rutgers, if you cannot wait.

	-- Bill Nowicki
	   Sun Microsystems