Re: [dnsext] Some thoughts on the updated aliasing draft

[Mark Andrews <[e-mail]>]

In message <[e-mail]>, "John Levine" writes:
> >if we're allowed by the people asking for this to require dns client
> >libraries and applications to have to be changed to handle a new kind
> >of canonicalization before they'll be able to handle the new
> >definition of "sameness" then this is a practical approach.
> 
> It seems to me that we can divide applications into two groups.  In one
> group, the client uses the name to find a server, but the name never
> appears in the data stream or only appears as a comment.  This includes
> FTP, SSH, POP, and IMAP.  Something like BNAME would be adequate to make
> names look the same to users.  I'd say these applications don't know
> what their names are.
> 
> In the other group, the domain name appears in the data stream, and
> the server does something with it beyond resolving it to an A or AAAA
> record.  The standard examples are SMTP and HTTP.  These apps do know
> what their names are, and the only alternatives I see are a great deal
> of manual provisioning, or else they learn how to check the DNS to see
> whether a hitherto unknown name in the data stream is an alias of a
> canonical name that it does know.  So, yeah, the applications will
> eventually have to change.

And SMTP had it correct for the second group.  If the SMTP client
sees the CNAME it re-writes the names in the SMTP exchange to use
the cannonical host name.  The SMTP server never see the alias.

HTTP administrators often miss use CNAME.

"www.example.net CNAME example.net" where the same content is
returned regardless of the the presense of www in the HTTP
stream is correct usage of CNAME.

If you use CNAME and you get different content then you are miss
using CNAME.  It's not being using to find the cannonical name for
the host.

Whether it is a virtual host or a name virtual host, the target of
the CNAME record is not the cannonical host name of the (named)
virtual host.  It is the hosting server for that site and that is
a very different thing.

> >From the point of view of DNS design, I'd want something that returns
> an appropriate A or AAAA for the first group, and offers a
> straightforward way for the second group to see what's an alias of
> what.
> 
> For a transition, it's still possible to configure the second group
> manually, but it'd be a lot easier if they got smarter so people
> didn't have to do so.
> 
> R's,
> John
-- 
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: [phone]                 INTERNET: [e-mail]