[99] in Hesiod
Re: Creating a new dns class
daemon@ATHENA.MIT.EDU (treese@crl.dec.com)
Fri Sep 4 17:21:06 1992
To: joann@cs.wisc.edu (Joann J. Ordille)
Cc: hesiod@Athena.MIT.EDU
In-Reply-To: Message from Joann J. Ordille of 4 Sep 92 15:59:03 CDT
Date: Fri, 04 Sep 92 17:18:30 -0400
From: treese@crl.dec.com
>
> I see from the current discussion that creating a new dns class
> has had it's ups and downs. From the spec (rfc 1035), this sounds
> like an easy thing to do, but the recent discussion indicates otherwise.
>
> Could someone summarize the pros and cons of creating the "HS" class for me?
>
> In particular, I'm interested in issues like the following ones.
> What's involved in creating a collection of "HS" class root
> servers and having them be known throughout the internet? Are there special
> interaction problems with the "IN" class? Aren't the trees completely
> independent? Does the bind resolver automatically support new and
> independent classes?
Creating a new class is far more trouble than it's worth. In the case of HS,
we shouldn't have done it.
Getting root servers is virtually impossible. Most DNS software (notably BIND)
doesn't handle different classes very well, and there isn't widespread agreement on
how to handle some of the problems that arise.
Win Treese Cambridge Research Lab
treese@crl.dec.com Digital Equipment Corp.
