2 more questions: domains and tags

[Phil Howard]

On Sun, Feb 02, 2003 at 09:39:53PM -0500, Ken Murchison wrote:

| Phil Howard wrote:
| > 
| > On Sat, Feb 01, 2003 at 10:12:52PM -0500, Ken Murchison wrote:
| > 
| > | Quoting Phil Howard <phil-info-cyrus at ipal.net>:
| > |
| > | > On Sat, Feb 01, 2003 at 11:34:24AM -0500, Rob Siemborski wrote:
| > | >
| > | > | On Sat, 1 Feb 2003, Phil Howard wrote:
| > | > |
| > | > | >     Is this done so transparently with Cyrus-IMAP that it didn't
| > | > | >     even need mention?  Or is it not done at all?
| > | > |
| > | > | There are some unofficial hacks to do virtual domains in 2.1, there's
| > | > | reasonable support for it in 2.2.
| > | >
| > | > How will bob at example.com and bob at example.net login?  Will it know which
| > | > user is logging in by them providing that as their login name?
| > |
| > | Yes.  The virtdomain support in 2.2 can use either fully qualified useridz (as
| > | you show above) or it can do a reverse lookup of the interface that the
| > | connection comes in on (for those who have IPs to burn).
| > 
| > Any idea when 2.2 will make it to BETA, or if a multiple-domain (only)
| 
| _Probably_ sometime this month.

Can you tell me if this is implemented by changing mailbox names adding
the domain name, or if it just simply keeps each domain in a separate
file tree much as one might do by having run each domain with a separate
daemon instance listing to specific interface addresses?


| > patch is available for 2.1.11?
| 
| I can pretty much guarantee you that no "official" virtdomains support
| will be added to 2.1.  In fact, the 2.1 branch is in a feature freeze.

One problem I often encounter in dealing with various mail programs
is this term "virtual domain" which I presume you mean by "virtdomain".
The problem is there are various different, and conflicting, overload
usages of this term.

One way "virtual domain" has been used is when an MTA (traditionally
sendmail) would accept any of several different domain names, but each
would be delivered as if the RHS was the same as the local machine.
This method required non-overloaded LHS addresses.  While many people
did use other terms, many did use "virtual domain" as well.

Another way allowed overloaded LHS by translating some or all addresses
to specific local users.  Again, there was a mix of usage.

The Apache web server uses the term "virtual domain" in several ways,
but generally being applied to how domain names and server names are
looked up or defaulted based on the destination IP address of the request.
In some configurations, each domain requires a different IP address.  In
others, they can overload the same IP address (HTTP 1.1 "Host: " attribute
required to distinguish domain names).

Having been in discussions, both for mail server issues as well as for
web server issues, where the term "virtual domain" was used in different
ways, sometimes 2 or 3 different ways by other participants, which
resulted in much confusion, often gnashing of teeth, and the occaisional
unmerited flamefest, I hope you can see that I am sensitive to any use
of the term "virtual domain".  It is one I wish would be banished from
any use.

In a case where the domain name is obvious, either because the connected
to IP address is associated with it, or the user logins in with it, and
is retrieving email which is distinguished by having been addressed to
the full user at domain which is logged in, then I would say there is nothing
"virtual" about the domain at all.  It is in fact very real.  What might
be considered special is that this domain isn't the same as the canonical
hostname of the machine serving the mail.  Using the term "virtual" just
because of this can lead to confusion because someone might think that
one of the other meanings may be intended.  I believe a better term for
such a domain is "hosted" or maybe "guest".  For example if the machine
is called "mail1.example.net" and I login as "bob at sales.example.com",
then "mail1.example.net" is the host or 'hoster', and "sales.example.com"
is hosted by it, and is thus the 'hostee'.

I'm certainly in no position to say which terms should be used beyond
just giving my suggestions.  But I really would urge not using the term
"virtual domain" without at least including a precise definition that
can exclude all other meanings besides the one intended.  But I think
we can find some better terms that will all be different for each of
the different concepts.  The choice of these terms might depend on just
how Cyrus 2.2 will implement the domains (I do not know this, yet).

-- 
-----------------------------------------------------------------
| Phil Howard - KA9WGN |   Dallas   | http://linuxhomepage.com/ |
| phil-nospam at ipal.net | Texas, USA | http://ka9wgn.ham.org/    |
-----------------------------------------------------------------