simple directory

Martin Hamilton (martin@mrrl.lut.ac.uk)
Wed, 20 Dec 1995 11:18:39 +0000


Message-Id: <199512201118.LAA10734@gizmo.lut.ac.uk>
To: uri@bunyip.com
Subject: simple directory
Date: Wed, 20 Dec 1995 11:18:39 +0000
From: Martin Hamilton <martin@mrrl.lut.ac.uk>

Here's an idea which seems to be relevant to URNs...

URLs (and most of the proposed URN schemes!) are dependent to a 
greater or lesser degree on the DNS: domain name syntax, character 
set restrictions, and so on

There's probably a fair bit to be gained from having object 
identifiers which aren't as closely tied to the DNS.  A couple of 
examples of this sort of identifier that are deployed already would 
be X.500's Distinguished Names and User Friendly Names

OK, X.500 is perhaps a tad too big and chunky for this sort of 
application.  There are lots of other directory service type 
protocols which could probably also be used - e.g. WHOIS++, RWHOIS or 
standalone LDAP

Arguably, the bare minimum of info which really needs to go into the 
directory service is the publisher's name and a domain name or names 
to look up for more info.  A directory which just consisted of 
publisher<->domain name mappings would seem to have better 
performance prospects (despite being distributed and global in scope) 
than one which contained arbitrary other stuff, such as White Pages 
info

Is anyone here is interested in working with some of the existing 
directory protocols to see if this approach is feasible ?  NB: I'm 
not suggesting that this be part of any IETF URI/URN/... effort, at 
least not initially.  If it turns out to work well, only _then_ do we 
start writing the RFCs :-)

I've made a bit of a start already -

The Netfind system developed by Mike Schwartz et al uses a "seed 
database" of organization names, domain names, and hosts which is 
derived semi-automatically from a variety of sources.  Here's a 
sample entry (wrapped):

  %D ecmwf.co.uk
  %O european centre for medium-range weather forecasts,
       reading, england, united kingdom
  %H unicos helena asclepius charybdis scylla xerxes jocasta
       agamemnon dido tantalus medea zephyr odysseus aeneas
       munin fulla loki www

It's not a complete record of every organization or Internet domain, 
and some of the info is out of date or just plain wrong.  
Nevertheless, I think it will do for test purposes

Since I already play around with WHOIS++ a bit, I thought to get the 
ball rolling I would try converting some of the Netfind info into 
WHOIS++ templates and indexing them. You can find the result at 
egate.lut.ac.uk, port 6663

Now, my server has just got the .uk info in it.  It would be great if 
a few other people could run up WHOIS++ servers with the info for 
(say) their top-level domain, and then we could try connecting these 
together in a mesh. I've generated templates for every top-level 
domain represented in the seed database - you can find these at 
<URL:http://www.roads.lut.ac.uk/~martin/wip/simple-directory/templates
/>.  Again, the data quality may be somewhat lacking, but I don't 
think this is an issue

You should be able to load these templates straight into Digger, the 
Bunyip WHOIS++ server.  More on Digger at <URL:http://services.bunyip.
com:8000/products/digger/>

I've set up a mailing list for people who want to write code, run a 
server, etc...  To subscribe - send mail to "deploy-request@mrrl.lut.a
c.uk, with a message body consisting of the word "subscribe" alone

Cheerio, (and Merry Xmas :-)

Martin