To: uri@bunyip.com
Subject: URN resolution needs to be reliable
Date: Mon, 02 May 1994 01:47:51 -0400
From: John Curran <jcurran@nic.near.net>
Message-Id: <9405020147.aa21000@nic.near.net>
Hmm. From reading the last few days worth of uri mail, it appears that
there's still significant diversity in what our requirements are for a
URN resolution service.
I'd like to explicitely mention my one principle requirement, so that I may
gauge support for this view (and identify any opposition for targetting :-)
I strongly believe that URN resolution should be *reliable*
By this, I mean that the resolution service should be robust, with
multiple servers (e.g. specified by multiple DNS records), using a
mature recovery technique (i.e. running over TCP or prospero/ARDP,
not UDP and random policy timers), and using a protocol which allows
the return of tagged information so that new elements may be added
in the future.
If we're not going to make URN resolution this good, we will never get
people to rely upon them. (To this day I'm still fighting with folks
who insist on using IP addresses because they're afraid DNS will fail
on them.)
/John