Re: Question for DNS propronents?

Michael Mealling (Michael.Mealling@oit.gatech.edu)
Thu, 15 Jun 1995 13:10:02 -0400 (EDT)


From: Michael.Mealling@oit.gatech.edu (Michael Mealling)
Message-Id: <199506151710.NAA24877@oit.gatech.edu>
Subject: Re: Question for DNS propronents?
To: moore@cs.utk.edu (Keith Moore)
Date: Thu, 15 Jun 1995 13:10:02 -0400 (EDT)
Cc: Michael.Mealling@oit.gatech.edu, uri@bunyip.com, moore@cs.utk.edu
In-Reply-To: <199506150505.BAA22451@wilma.cs.utk.edu> from "Keith Moore" at Jun 15, 95 01:05:22 am

Keith Moore said this:
> > How do you deal with information spaces and server hierarchies that
> > don't correspond to actual DNS name/server hierarchies?
> 
> One way is to define them in other places under the DNS root.  There's
> no particular reason that your naming authority name has to be in the
> same DNS subtree as your host names.  And if you want your URNs to
> last awhile, there's good reasons not to do this.

Ok. This I can live with. Many of the previous suggestions was that
we stick with the existing namespace AND NAMES. I.e. I was stuck
with gatech.edu. Basically any '.' delimited most->least significant
order naming system is compatible with DNS as long as it doesn't
contain very strange characters.

> There are about three separate questions to "should we use DNS"?
> 
> 1) should we use the existing DNS hierarchy?
>    (e.g. we could restrict all URN naming authorities to a particular
>     subtree of DNS, or we could define one or more new subtrees for 
>     "persistent" URN naming authorities, while allowing people to
>     use any DNS name if they want to)

I agree. Information hierarchies don't necessarily work well bound 
withing subdomain/machine hierarchies. And both shouldn't have to be
kludged for  the other...

> 2) should we use DNS query protocols?
>    if we don't like the way dns exposes its naming hierarchy, 
>    (i.e. that the resolution hierarchy is inseperably wired into
>    the dns name) we can build our own DNS servers for particular
>    subtrees of DNS name space and they can resolve names any way
>    they want to (like iddd.tpc.int)

I think we should simply because they have been written generically
enough to be used for just about anything. One proposal I had
was to setup a seperate DNS server hierarchy on ANOTHER PORT. In
this case its 4053 so that it can be public. Its very easy to have
libresolv talk to another port. I'm currently hacking a small
C interface that gives you the equivelent of gethostbyname but
for OIDs (see Larry's latest mail).

> 3) should we use DNS servers?
>    if named isn't reliable enough or secure enough or doesn't
>    permit the kind of updating needed by a particular naming
>    authority, they can build their own servers.

named is reliable as long as you don't make the same process do hostnames
URNs, OIDs, ISDN names, etc. Break each of these into its own server
tree running on different ports so one system doesn't cause load on
the other.


> To DNS or not to DNS?
> I'm still not sure which approach is best, but I haven't seen
> any objection to DNS that can't be dealt with.

I used to object but my scenario is hard to complain against. I talked
to Paul Vixie about it and he likes it and can't see any problems.

-MM

-- 
------------------------------------------------------------------------------
Life is a game. Someone wins and someone loses. Get used to it.
<BR>
<HR><A HREF="http://www.gatech.edu/michael.html">Michael Mealling</A>