RE: Confusing the DNS (was: Re: [wg-c] Straw Vote)

[Dave Crocker <dcrocker@brandenburg.com>]

At 07:43 AM 8/17/99 , Roeland M.J. Meyer wrote:
>I both agree and disagree. DNS isn't perforce hierarchical, as dlbDNS
>seems to indicate. It is a configuration issue. Karl and I have been

No it is not a "configuration" issue.  The technology is fully and 
inherently hierarchical.   The fact that each user of the DNS can choose to 
use a different DNS does not change the nature of the technological base.

ISPs and others choose which hierarchies they will point to.  However each 
thing they point to IS hierarchical.

>talking about multi-part roots and heuristics involved in building a
>root search capability into BIND. This also answers the reasons that

Talking is easy and often is misleading.  Discussion about the feasibility 
of multiple roots both flies in the face of expert technical analysis and 
in operational history.  In other words, for all the talk, we haven't seen 
it happen yet.

It is generally considered risky to make major design, configuration and/or 
operations changes to a running piece of critical infrastructure 
service.  To make such changes in the absence of solid, empirical 
experience is far worse; at the most polite it would be called imprudent.

>there are only 13 root servers (which was an architectural performance
>issue, on a VAX, and not a technical limitation of the code, BTW). After

Roeland, the limit has nothing at all to do with using a Vax, or any other 
machine dependency.  It is dictated by operational size limitations of 
UDP.  In other words, it is a protocol limitation, not a machine limitation.

d/

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Dave Crocker                                         Tel: +1 408 246 8253
Brandenburg Consulting                               Fax: +1 408 273 6464
675 Spruce Drive                             <http://www.brandenburg.com>
Sunnyvale, CA 94086 USA                 <mailto:dcrocker@brandenburg.com>