[comp.protocols.tcp-ip.domains] Creating a domain alias

emv@ox.com (Ed Vielmetti) (03/25/91)

In article <KIMMO.SUOMINEN.91Mar24234231@kannel.lut.fi> Kimmo.Suominen@lut.fi (Kimmo Suominen) writes:

   Is it really correct to have two names pointing to the same address.
   The way I see it, one of these names should be a CNAME entry. 

CNAMEs are 2d class internet citizens.  Many things are not legal when
CNAMEs are around, but when the suitable A records are put in their
place things work.  It is (e.g.) illegal to send mail with a 
	From: cname.domain.org
that is to say the From: line must include an MX or an A record.  Put
in an A record for the domain and the host in the domain and all will
be well.

-- 
 Msen	Edward Vielmetti
/|---	moderator, comp.archives
	emv@msen.com

Kimmo.Suominen@lut.fi (Kimmo Suominen) (03/25/91)

[ I am moving this conversation to comp.protocols.tcp-ip.domains,
  where it belongs better, IMHO ]

>>>>> On 23 Mar 91 23:54:15 GMT, sidney@borland.com (Sidney Markowitz) said:

Sidney> The simplest way is to add an A record that defines the domain name as
Sidney> a host with the same IP address as the mail host. Here's what it looks

Sidney> $ORIGIN borland.com.
Sidney> @		IN	NS	frank.borland.com.
Sidney> frank		IN	A	131.119.250.105
Sidney> $ORIGIN com.
Sidney> borland		IN	A	131.119.250.105

Now, IS THIS CORRECT?

Is it really correct to have two names pointing to the same address.
The way I see it, one of these names should be a CNAME entry.  And
because a domain cannot be a CNAME, then the name "frank" should be a
CNAME ponting to borland.com.  Like this

	$ORIGIN borland.com.
	@	IN	NS	@
		IN	A	131.119.250.105
	frank	IN	CNAME	@

Of course this means, that the canonical name for the host is then
borland.com, not frank.borland.com.  But I can't see, what the
canonical name is in the original case...

Please, comment on this and let me know, what the right way is.  I
think this has been discussed before, and I think, that the way I
described is the correct one.  Am I right?  Or am I right? ;-)

P.S.  I am not trying to offend anyone.  I have just run into this
problem many times lately.
--
Kim                      /  Internet: Kimmo.Suominen@lut.fi
"That's what I think."  /   Bitnet:   KIM@FINFILES

Kimmo.Suominen@lut.fi (Kimmo Suominen) (03/25/91)

>>>>> On 25 Mar 91 02:15:50 GMT, emv@ox.com (Ed Vielmetti) said:

Ed> CNAMEs are 2d class internet citizens.  Many things are not legal when
Ed> CNAMEs are around, but when the suitable A records are put in their
Ed> place things work.  It is (e.g.) illegal to send mail with a 
Ed> 	From: cname.domain.org
Ed> that is to say the From: line must include an MX or an A record.  Put
Ed> in an A record for the domain and the host in the domain and all will
Ed> be well.

Well, it is not recommended to use an alias in mail addresses.  But I
don't see why it wouldn't work?  You CAN mail to the following
addresses (which don't have a MX-record):

	Kimmo.Suominen@kannel.lut.fi
	Kimmo.Suominen@news.lut.fi
	Kimmo.Suominen@ftp.lut.fi

They are all CNAMEs and the mail will be directed to the canonical
name lut.fi.  So far I don't see a problem here...
--
Kim                      /  Internet: Kimmo.Suominen@lut.fi
"That's what I think."  /   Bitnet:   KIM@FINFILES

cliff@garnet.berkeley.edu (Cliff Frost) (03/26/91)

In article <KIMMO.SUOMINEN.91Mar24234231@kannel.lut.fi>, Kimmo.Suominen@lut.fi (Kimmo Suominen) writes:
...
|> 
|> Is it really correct to have two names pointing to the same address.

Why not?  It may be harder to maintain consistency in your database, 
but if you want to do it I don't see why not.

|> The way I see it, one of these names should be a CNAME entry.

Imagine a host with multiple interfaces.  You might like to be able
to say "telnet host" and have telnet try all the interfaces.

But you might also like to be able to say "telnet host-ether" and have
telnet try only one of the possible interfaces.

Imagine you have services running on multiple machines (maybe some have
multiple interfaces).  Your nameservers might be NS.domain, your FTP
repositories might be FTP.domain, etc.

If you want to control which machines/interfaces get used for these
services you may have to assign addresses to each service-name.

I believe that what is restricted is that each ADDRESS can have only one
PTR record.  That is, each ADDRESS has one canonical name.  But, many
names can use the same address.

	Cliff Frost
	UC Berkeley

gandalf@csli.Stanford.EDU (Juergen Wagner) (03/26/91)

Kimmo.Suominen@lut.fi (Kimmo Suominen) writes:
>Well, it is not recommended to use an alias in mail addresses.

In some cases, it is *desirable* to have CNAMEs in mail addresses. If a site
with a large number of machines and users decides to promote a single,
uniform addresssing schema for all users to the rest of the world, there are
three alternatives:

-  mailbox@somehost.our.domain	(somehost being a real host)

   This has the disadvantage of having to be changed when the mail service
   is moved to another machine.

-  mailbox@our.domain		(our.domain being the local domain name)

   Having MX for our.domain point to somehost.our.domain, this is a general
   solution. The MX host can decide what to do with mail on a local basis.

-  mailbox@virtual.our.domain	(virtual being a non-existent host name)

   Entering MX or CNAME records for the virtual host provides the same
   functionality as the second solution. Users have to type more, though.

As the third example shows, CNAMEs provide an easy way of calling hosts by
more than one name. This allows people to call hosts not only by their
canonical name, but also by abstract names designating some service or
function (e.g. ftp.iao.fhg.de, nntpserver.iao.fhg.de, ...).

As for having multiple names pointing at the same address, I am not sure
whether that's recommended. However, I found it useful in situation where
hosts have multiple interfaces, so I can refer to the interfaces by distinct
names:

apatix		IN	A	129.233.6.1
		IN	A	192.44.21.1

apatix_qe0	IN	A	129.233.6.1
apatix_dmv0	IN	A	192.44.21.1

with all addresses pointing to the canonical host name 'apatix'.

--Juergen Wagner
(gandalf@csli.stanford.edu, a.k.a. wagner@iao.fhg.de)