Re: URNs in the DNS

Martin Hamilton (M.T.Hamilton@lut.ac.uk)
Fri, 16 Jul 1993 13:23:05 +0100 (BST)

Date: Fri, 16 Jul 1993 13:23:05 +0100 (BST)
From: Martin Hamilton <M.T.Hamilton@lut.ac.uk>
Subject: Re: URNs in the DNS
To: mitra@path.net
In-Reply-To: <CA83zJ.BrC@pandora.sf.ca.us>
Message-Id: <Pine.3.07.9307161305.E5963-b100000@lust>

Mitra said:

> Every DNS type person here I've spoken to thinks this would be a bad
> idea since the DNS needs to be optimised and really fast for what it
> does, adding two orders of magnitude mor information to it might not
> break it, but it would hardly then be optimised for turning FQDN's to IP
> addresses.

Indeed! If we inadvertantly broke the DNS while playing with
some hare-brained new scheme there'd be a lot of angry people
out there...

Nevertheless, converting symbolic names to "physical" addresses
is the job we currently ask the DNS to do for us, and it isn't
stretching the point *too* much to ask it to do the same thing
for URLs. I think the main questions would be what happens in
terms of scaling and caching - e.g. TTLs too small and you're
forever repeating lookups on popular objects, TTLs too big
and the servers fill up with gunk. Too much gunk in your
server and it stops working - ouch!

Personally, I favour the general purpose directory approach,
but the DNS based alternative is fascinating, isn't it ?!?
I *had* to suggest it!

Thankfully we have X.500 to play with - "where no-one can hear
you scream..." :-D

Martin